Engineering the Harness
The only way to trust a knowledge base is to test it. So I did, and found that almost nothing it told me was wrong. Just thin.
Insights on AI, document intelligence, and workflow automation.
The only way to trust a knowledge base is to test it. So I did, and found that almost nothing it told me was wrong. Just thin.
I built my AI agents a knowledge base, then realised the real trick was keeping it out of context until they need it — and that a line I'd written for tidiness had quietly become a routing table.
The ralph loop that changed how I build.
What I learned integrating an AI agent into a legacy Windows Forms application.
Reviewing what happened last time so I can improve this time.
Why I'm scrapping my first AI-generated MVP and starting again.
Robot Brains is my AI Lab—learning through doing, and maybe creating something great along the way.
Our first post - introducing the robotbrains blog and what to expect.