Field notes

Short notes from ideas I am still working through. Some came from older essays. Some came from recent conversations. They are here to be challenged, not treated as finished.

This site is built to be read by your agent, not just you.

Field notes

I don't think at a blank page, I think at a sparring partner

I knew I was a translator before I had the words for it

I measured where my AI tokens actually go

Brand was the sending half. AI is the receiving half.

Show, don't tell never sat right with me

Catch the spark, it has a location

It has to live where you already think

You're always choosing which branches to render

Diverge to find the boundaries, converge to render

Give your agents a living manual

The multiverse in reverse

We chain sparks until they catch

A compass, not a map

AI memory is flat, and that's the problem

Calibrate, synthesize, communicate

Make it possible to pull the threads

Selling my brain with the outcome of a logo

Don't let AI flatter you

Show the work, because the work is the proof

Texture is what sets your audience's posture

The secret isn't the recipe, it's that they won't do it

Why you feel disconnected when someone won't match your posture

The thinking space should stay light

A summary can be accurate without being true

Collaboration over empires

All good thinking comes from wrestling with low-confidence ideas

Orientation matters, but shared understanding matters more

Prompt anxiety is the new performance anxiety

The map needs a destination

The words are the costumes

We're compounding data, not thinking

Low-confidence thinking needs a license

Keep the record useful for the next decision

A fresh thread should not mean starting over

Arrive with curiosity, not a conclusion

Done means proof, not a confident claim

Context should move through valves, not walls

Don't describe the view from the hill, help people climb it

The conversation is the artifact

The discipline is in the dosage

You can't read the signal through flattery

Reading intent, not just the message

Volume or conversion, which do you fix first?

Naming the problem vs. ordering the solution

Does an audience sharpen your thinking or soften it?

We keep asking AI to be an engineer when it's better as a reasoner

AI-generated design has tells

My taste is better than my skills

The discipline I keep skipping is the swipe file

When everyone can vibe-code their own tools, what's left to sell?

An agent that looks like it did research usually didn't

Don't build the $100M system. Build the thing that gets you to next month.

Financial modeling and business modeling are two different things

A fitness app should be a companion, not a coach

Build for the most technical, default to the simplest

The build loop: from brain dump to shipped while you sleep

Why so many internal AI tools quietly break

The human decides what to build. Agents decide how.

Get the function working before you fall in love with the design

Stop forcing two things to be one thing

What if the company building the product was mostly agents?

It's not your job to teach people to use AI, build them the tool

Collecting tools isn't the same as internalizing them

Editing with AI should feel like working on an artifact, not starting over

The best plan feels like it lives with you

The promise has to hold with your favorite feature turned off

Design the primitives before the screens

Don't let the bad days stack

Fitness has to fit the life you actually live

In the real moment of use, everything extra is noise

Team context and personal context need a membrane

Simple principles, personalized dosage

A health app should work with the rhythms of your life, not against them

The next big gap between companies is who builds agent teams

AI shouldn't placate you, it should challenge you

An organization needs an evergreen layer, not just files

How far is a working prototype from a real product?

Should you build software for agents as much as for people?