Recent Posts
-
How I run agents in corporate codebases
A corporate feature spans many repos. I run the agent one level up, from a workspace that holds all of them, with a map of how they connect and a docs folder it researches into.
-
OpenClaw: the Mac minis you don't need
You don't need a stack of Mac minis to run multiple OpenClaw agents. Sandboxes give you the isolation people are buying hardware for, on a single machine.
-
Stop Making Your AI Agent Click Buttons
Why I stopped using LLMs to click through websites and started writing JS adapters that call the data layer directly.