I build software, follow questions, and write about what I find.
Systems, AI, design, and the occasional loose thread.
A working notebook,
by Bri Stanback
Better questions
compound.
Latest essay
· 24 min read
Sameness isn't one failure, it's two — a model's own training narrows what it's willing to produce before anyone reviews anything, and a review loop narrows what survives on top of that. Same pressure, two different rooms — and I've built something against both, though I don't yet know if either one actually holds.
Read the essay ↗︎A Notion co-founder says everything he knew about coding agents six months ago was wrong. Most of what changed validates an older idea: constraints do the work, not agents — but at multi-agent scale, the constraint that matters most is who holds which decision rights.
The interesting part is the fractal — tight loops inside medium loops inside long loops, each with different primitives and different levels of human involvement.
Everyone wants to be the source of truth. Almost nobody builds the structural conditions that make truth survivable. From the Delphic Oracle to Grokipedia, the pattern is the same — and it has nothing to do with intent.
Every system has defaults. AI makes them more powerful and more invisible. The question isn't whether your system is biased — it's whose assumptions it's running on.
Software engineering compensation was built on scarcity, not just difficulty. That scarcity is eroding. The people who survive are either unusually broad or unusually deep.
Vector search in the 90s, badges that followed you through the office, and what happens when you build an AI system without knowing what's already been built.
A neuromarketing talk got me thinking about reward systems, GLP-1s, and what happens when AI becomes the most sophisticated rationalizer we've ever built.
Short, visual explorations of ideas I’m thinking through.
Longer-form discussions with people I respect.
