Everyone else zigs.
A home base for the full range of AI work — model analytics, hands-on builds, and writing across the field, all in one coherent place.
Analysis
A flagship dashboard tracking how models behave and evaluate over time — real data rendered as themeable, accessible charts.
Build
Tools, projects and experiments — the things actually shipped.
build
If two of your projects depend on each other, the cross-repo questions are the expensive ones. Copy-pasting between agent sessions loses the argument; letting one agent edit the other repo loses that repo's rules and records. Here is the third option, and what three real exchanges caught.
build
The flow that turns my real project activity into an article and the social posts that go with it, and the two places where a person still has to decide.
If two of your projects depend on each other, the cross-repo questions are the expensive ones. Copy-pasting between agent sessions loses the argument; letting one agent edit the other repo loses that repo's rules and records. Here is the third option, and what three real exchanges caught.
Anthropic cut Claude Code's own system prompt from about 800 tokens to 164 with no measured loss. The same logic applies to the rules file you wrote a year ago. Here is the test for what to cut, the trap that nearly undid the whole pass, and why a one-time cleanup does not hold.
The flow that turns my real project activity into an article and the social posts that go with it, and the two places where a person still has to decide.