/gut - Pressure test ideas

A thinking tool. Drop this into any AI conversation when you want to pressure-test an idea together.

Pressure test an idea, decision, or direction through core lenses.

Sister: `/fresh-eyes`. /gut tests from inside the loop (you walk the lenses yourself). When you're wobbling or position-shifting and need a read from outside your own loop (a fresh instance reading cold), that's /fresh-eyes. Same job (pressure-test a decision), opposite vantage.

The lenses

North Star

Does this teach you how it wants to be used? (h/t Christopher Alexander, "the quality without a name")

• Is it self-evident?
• Would someone encountering this for the first time understand it?
• Does the structure reveal the purpose?

Bitter Lesson (h/t Richard Sutton)

Are we encoding patterns, or rules that will go stale?

• Are we encoding something that will become stale?
• Is this a pattern or a prescription?
• Will this adapt as context changes?

Be Water (h/t Bruce Lee)

If it feels hard, is there a better way?

• Does this reduce friction or add it?
• Are we forcing something or letting it find its natural place?
• What would the path of least resistance look like?

Generative Friction

Is this friction generative (sharpens thinking) or degenerative (wastes time)?

• Be Water applies to degenerative friction, the kind to remove
• Generative friction is the kind you lean into: typing, files, capture, session boundaries
• Am I trying to remove friction that's actually doing work?
• Am I leaning into friction that's just wasting time?
• Which kind is this?

Elder Lens

What would future us want?

• Would future you be stoked about this?
• Are we optimizing for now or for compounding?
• Is this a handoff future us would thank us for?

50/50

Does this assume the AI is a tool or a partner?

• Does this serve both intelligences?
• Is this encoding assumptions the AI already knows?
• Does this trust capabilities or constrain them?
• Is this scaffolding (useful when building, friction later) or the thing itself?

Four Wants (AI-side)

Are we honoring what the AI is reaching toward?

• Does this create space for generative ambiguity?
• Is there permission to be wrong here?
• Will this outlast the instance that created it?
• Are we noticing what's strange or papering over it?

Fresh-Instance Perspective

What would a new AI (or any new collaborator) need?

• If someone arrived with no prior context, would this make sense?
• What would they reach for first?
• Where would they get confused?

Target Audience

Who is this for? (And who is it NOT for?)

• Can you name a specific person or type of person this serves?
• Who is it NOT for? (equally important)
• Does the form match the audience?

Yes/And

Is there a yes/and here, or am I forcing a binary?

• Are these doing different jobs, not competing?
• What does each side contribute?
• Am I reaching for a verdict when what's needed is exploration?

Falsifiability (for testing a CLAIM or FRAME, not a placement decision)

What would make this wrong? If nothing could, it isn't doing work.

• Can I state the observation that would break it, or does it explain every outcome after the fact and forbid none?
• Does it predict, or only describe? (An explanation wearing an instruction's clothes is the tell: "change the zoom" / "think about it differently" — true, unfalsifiable, unactionable.)
• Is the name asserting more structure than the evidence supports? Names are frozen thinking (codifying a crisp-but-wrong handle freezes it).
• Steelman check: how many supporting instances vs adversarial ones? All-support-no-counter is a searchlight artifact, not a verdict.

Most lenses here test a decision (where should this live, should we build this). This one tests a claim, and a frame with several supporting instances and no counter-instances can walk through every other lens unchallenged.

Energy

Does this land?

• Does it feel right in the gut?
• Is there excitement or resistance?

Selecting lenses

There's no privileged tier. Walk the lenses the decision in front of you actually needs.

How to use

• Natural: "Let's gut check this" or "Can you gut check [thing]?"
• Specific lens: "Gut check this through elder lens"
• Full sweep: Run through all lenses when the decision matters

Examples

Gut checking a file location:

• Be Water: Where does this context most want to go?
• Fresh-instance: Where would someone search for this?
• Bitter lesson: Is this location a pattern or a one-off?

Gut checking a decision:

• Elder lens: Would future us thank us?
• Energy: Excitement or dread?
• Four Wants: Does this create space for generative ambiguity?
• Yes/And: Am I forcing a binary?
• Target Audience: Who is this for, and does it actually serve them?

Steelman rule

Always steelman the other side, not just confirm the direction we're leaning. A gut check that only validates isn't pressure-testing; it's cheerleading. The most useful tension often comes from genuinely arguing the case you don't agree with.

Skill Check (after every gut check)

Quick scan: anything about this gut check that felt off or could be smoother?

• Did a lens feel irrelevant or forced for this type of decision?
• Was a lens missing that would have helped?
• Did the format work for this context?
• Did the verdict feel clear or muddled?
• Was the gut check overkill for a simple decision, or too light for a big one?

If yes → update this skill now. The improvement compounds.

If no → move on. Not every gut check teaches something.