5 min read
Recipes don't work for AI agents as well as you think
A one-line cultural value in my AI agent's prompt outperforms 200 lines of rules. Principles generalize across edge cases; recipes don't.
writing
5 min read
A one-line cultural value in my AI agent's prompt outperforms 200 lines of rules. Principles generalize across edge cases; recipes don't.
8 min read
Founders treat onboarding as a building discipline when it's an editorial one. Cut until what's left is the smallest experience that still hooks the user.
5 min read
The next venture-scale AI wave isn't limited to the bottom of the labor pyramid. It includes credentialed but tedious work like bookkeeping and legal due diligence.
7 min read
Two AI apps running the same models need wildly different accuracy rates to be trusted. What sets the bar is how much work the user does to verify each output.
5 min read
When the consensus rejects your idea for a wrong reason or assumption, the rejection is positive signal to pursue it. The hard part is naming the specific assumption everyone else is missing.
5 min read
Your unconscious filters out startup ideas that look like tedious work (schlep blindness). AI in 2026 just made every existing schlep cheaper to solve.