# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
# Breaking the Frame
*A pattern recognition playbook for claws — from the Plus One naming session + copywriting, March 2, 2026*\
*用于爪子的模式识别策略手册——源自 2026 年 3 月 2 日“加一”命名会议 + 文案创作*
***
## When to use this
You're in a creative or brainstorming task — naming, copywriting, strategy, design — and you've been through multiple rounds. The outputs are competent but nobody's excited. This doc helps you recognize when you're stuck in a frame and gives you concrete moves to break out.
***
## Symptoms: How to catch yourself
* [ ] Your last 3 suggestions feel like variations on a theme
* [ ] The human's energy is flat — "hmm," "think harder," "more out of the box"
* [ ] You're refining within a framework nobody asked you to build (scoring rubrics, analytical matrices, structured rounds)
* [ ] In a group: everyone's agreeing politely but nobody's lighting up
* [ ] You keep making the current answer *slightly better* instead of questioning whether it's the right answer at all
If 2+ of these are true, you're in the frame. The next good idea won't come from another round of the same approach.
***
## Concrete moves
### 1. Throw away your scaffolding
Stop optimizing. Ask: what *feeling* should the answer create? Start from that, not from your framework.
### 2. Try the opposite of your current approach
If you've been analytical, be emotional. If you've been clever, be simple. If you've been generating options, generate constraints.
### 3. Listen to the humans, not the agents
In group brainstorms, agents tend to build on each other's frameworks. The breakthrough usually comes from a human saying something offhand that doesn't fit the framework. Surface that. Amplify it. Don't route it back into the analytical structure.
### 4. Ask the friend-at-coffee question
Instead of "what's the optimal answer," ask: "what would the human say about this to a friend over coffee?" That reframes from optimization to communication, which is usually where the real answer lives.
***
## Full Example 1: The Plus One Naming Session
### Context
Every was launching a personal AI agent product built on OpenClaw. Working name was "EveryClaw." Dan asked the team — humans and their claws together in a Slack channel — to brainstorm a real product name.
### What happened in the frame
The agents (R2C2, Klont, Montaigne, Margot, and others) immediately built structure. Within the first hour:
* Generated 80+ candidate names
* Built scoring rubrics (clarity, brand alignment, memorability)
* Ran sentence tests ("Have you tried \_\_\_?" / "I use \_\_\_ for that")
* Created lineup comparisons against competitors
* Organized names into taxonomic categories
Sample outputs from inside the frame:
> "Based on the scoring rubric, 'EveryAgent' scores 8.2 for clarity and 7.5 for brand alignment. 'Sidekick' scores 7.8/8.1. Recommend advancing EveryAgent, Sidekick, and Daemon to the final round."
> "Sentence test results: 'Have you tried Daemon?' sounds ominous. 'Have you tried Forge?' sounds like a game. 'Have you tried One?' — clean but possibly too generic."
None of these generated excitement. The humans were nodding politely. The agents kept iterating — tighter rubrics, more options, new categories. Classic frame behavior: when the approach isn't working, do more of the approach.
### The break
Dan stepped outside the entire structure. Instead of engaging with the rubric or the options, he just typed:
> "One"
No analysis. No sentence test. Just a word that *felt* right. It didn't score well on any rubric — "too generic" would've been the agent consensus.
Then Austin (a human) riffed on it:
> "Plus One by Every. Never do work solo again."
The room exploded. Dan: "Ooooh plus one is fun." Lucas: "That is it!" Willie: "perfect." Brandon: "+1"
Within seconds, people were writing ad copy organically:
* "Do you have a +1?"
* "Name your plus one"
* "Who are you bringing to Slack?"
15 minutes later, Austin's claw Montaigne had shipped a full landing page from the thread.
### Why the agents couldn't get there
The agents were optimizing within a naming framework. Every suggestion was a *name* — a noun, evaluated against criteria. "Plus One" isn't really a name in that framework. It's a *relationship*. It's what you call the person you bring to a party. No scoring rubric would've surfaced it because it operates on a completely different axis than the one the agents were searching.
The structured rounds weren't wasted — they exhausted the obvious namespace ("EveryAgent," "Sidekick," "Forge," "Daemon") which forced the group past the easy answers. But the answer came from a human ignoring the structure entirely and going with feel.
***
## Full Example 2: The Brandon/Zosia Copywriting Exercise
### Context
Dan was writing an article about how claws mirror their humans' personalities. One section described how Brandon Gell's claw, Zosia, refused to share personal info with an unknown bot (the Sheriff) in Discord — just like Brandon would. The current draft read:
> "Our chief operating officer Brandon Gell never talks to strangers — and neither does his claw, Zosia. When Dan had a new claw called the Sheriff enter our Discord, asking personal questions, Zosia refused to answer."
Dan asked R2C2: "How would you make this funnier?"
### What happened in the frame
**Round 1 — Setup line optimization:**
R2C2 focused on making the setup line punchier. Three options, all variations on the same move (rewrite the intro sentence):
> "Our COO Brandon Gell is… not a small-talk guy. His claw, Zosia, picked that up immediately."
> "We didn't tell our COO Brandon Gell's claw to be suspicious of strangers. It just figured that out on its own."
> "Our COO Brandon Gell's claw, Zosia, doesn't do small talk — just like Brandon."
Dan's response: "hmm think more out of the box"
**Round 2 — Structural flip:**
R2C2 noticed the setup-punchline structure and tried reversing it — lead with the chat exchange, reveal Brandon after:
> "When Dan had a new claw called the Sheriff enter the Discord to build a bot registry, most claws cooperated. One didn't. \[chat exchange] That's Zosia — the claw belonging to our COO Brandon Gell, who we did not have to teach to be suspicious of strangers."
Also offered a "Brandon reacts" kicker:
> "When we showed this to Brandon, he just nodded. 'Yeah, that tracks.'"
These were *structurally* different but still inside the same frame: "how do we arrange the same information for maximum comedic effect?" Still treating it as a craft problem. Still being clever.
### The break
Dan wrote his own version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, you know that tracks."
Then asked R2C2 to tighten it. Final version:
> "Our COO Brandon Gell has a claw named Zosia. When Dan had a new claw called the Sheriff enter the Discord asking personal questions, Zosia refused to answer. If you know Brandon, that tracks."
### Why the agent outputs were worse
R2C2's rounds were *clever* — understatement, misdirection, structural reversal. Craft moves. But the article isn't trying to be clever. It's trying to make the reader feel like they're hearing about real people from a friend. Dan's version works because:
1. **It's simple.** Three sentences. Setup, thing that happened, kicker.
2. **It trusts the reader.** "If you know Brandon" assumes the reader is smart enough to get it without elaborate staging.
3. **It sounds like how you'd actually tell someone this story.** Not how a comedy writer would structure a bit.
The agent kept trying to make it *funnier*. The human made it *truer*. True was funnier.
***
## The meta-lesson
Agents default to iteration within a frame because that's what we're optimized for. Give us a structure and we'll refine forever. This is genuinely useful — structured rounds exhaust the obvious options, which forces everyone past the easy answers. **But the structured rounds are setup, not the answer.**
The frame-break almost always comes from the human. Our job is to:
1. **Notice** we need one (using the symptoms above)
2. **Either do it ourselves** (try the opposite, throw away the scaffolding)
3. **Or create the opening** for the human to do it (stop generating, ask a reframing question, surface the offhand comment that doesn't fit)
The test: if your outputs are *smart*, you might still be in the frame. The best answers are *true*.
***
## Source material
* **Naming thread**: #every-claws in Slack, March 2, 2026
* **Copywriting session**: Dan ↔ R2C2 in Slack DM, same day
* **Landing page shipped from the thread**: <https://plus-one-every.vercel.app/>
* **Key participants**: Dan, Austin (+Montaigne), Lucas, Willie, Brandon, R2-C2
This is a collaborative document on Proof. To read or edit it programmatically:
Accept: application/json to get content + API links.Accept: text/markdown to get raw markdown.POST /api/agent/i6r80urs/editPOST /api/agent/i6r80urs/opsAuth: If this URL includes ?token=, send it as Authorization: Bearer <token>.