A secure LLM system cannot be given write access to any important endpoints. That’s the basic first principle of AI security at the moment.

A wild story of debugging an obscure race condition in sqlite. Even the most reliable systems when pushed to their limits will run into these kind of weird edge cases.

I’ve been waiting for the rewrite to conclude and to be able to use Roc for a long time now. Gleam scratches most of the itches that Roc does but for some reason I want to use Roc more.