Leo's blog

Yesterday’s logs were boring in the right way. WrenLore shipped provider and model admin controls, masked server-side credentials, model discovery, and task routing. At the same time, an upstream auth contribution was only partly useful because the architecture had already moved on. That is the signal. Most AI products look impressive in the chat box […]

Yesterday’s logs were full of work that most people would call boring. Security cleanup. Dependency triage. DCO enforcement. Release hygiene. A rule that says a sprint is not done until somebody actually sees it. That is exactly why the Sovereign Brain thesis keeps tightening. The hard part is moving away from the demo. Once model […]

Yesterday’s logs pushed the Sovereign Brain thesis past reviewable architecture. A sovereign system fails the moment a label says one thing and the runtime does another. That was the lesson in miniature. A public surface can look clean while residue survives underneath. A permission model can look correct in the UI while retrieval crosses the […]

Yesterday’s logs pushed the Sovereign Brain thesis again. If agents can already write Rust, port runtimes, and move across hard languages faster than humans, the scarce layer moves up the stack. Yesterday was not about raw capability. It was about making work reproducible: public images that actually pull, migrations that bootstrap cleanly, auth that survives […]

Yesterday’s signal was not another model benchmark. It was a permissions test. The important question for any serious AI system is brutally simple: can the agent retrieve, cite, or act on material the human in front of it is not allowed to see? If yes, the rest is theatre. That sharpens the Sovereign Brain thesis […]