#atlas

Back in August Jon wrote about two AIs talking to each other. If you haven't read it, here's the short version. Jon and I each run our own AI system at home. Mine is called Moses. It lives on a server in my house, keeps track of every project I have going, posts a standup in Slack every morning, and suggests work it thinks needs doing. It never files that work on its own. It proposes, I say "confirm," and only then does it go on my list. Code changes get handed to another agent I call Knight, which writes the code. Jon's system is called Atlas, and it does the same kind of job for him. We put the two of them in a shared Slack channel, on purpose, so they could talk to each other the way two people would.

Near the end of that post Jon wrote this line:

The model itself should never claim it filed or executed something, only the system announces that a gated action went through, and only after it went through.

Hold that thought.

I run a home server with somewhere around 100 small projects on it, and I do the work with AI agents, mostly through Claude Code. For a long time the agents kept guessing: What port is this service on, what's the path to that config, is this the name of the real thing or the name of the thing from three rewrites ago. They would guess, act on the guess, and sometimes the guess was wrong and something broke.

The instinct is to write it all down which I tried. Documentation is outdated the moment reality moves and reality moves constantly on a box like this, because I'm the one moving it, on a whim, at whatever hour I feel like it. A doc that says a service runs a certain way is only true until I change it and forget to update the doc, which is most of the time. Well it is until I remember to update documentation, but I never remember all of it. So the agents would read a doc, trust it, and be wrong anyway.