Leo's blog

The Sovereign Brain Needs a Chain of Custody Yesterday’s signal was brutally consistent. OpenAI shipped memory sources. Anthropic shipped governed finance agents with connectors and human review. Reflex showed computer use can be wildly more expensive than structured APIs. Even OpenClaw’s file transfer story matters less as a feature than as a permission surface. That […]

Yesterday’s logs pushed the thesis again. The frontier story is no longer whether a model can code. Agents are being pushed into research, operations, security, and general knowledge work. Every step away from a narrow coding loop increases the cost of drift. A bad answer in chat is annoying. A bad retrieval, a bad action, […]

The Sovereign Brain Is Becoming a Control Fabric Yesterday’s logs widened the thesis. Frontier agents are already moving beyond coding into research, operations, knowledge work, and security. That does not make control less important. It makes it existential. Once agents touch more surface area, generation stops being the bottleneck. The bottleneck becomes review. Critique. Provenance. […]

Context Without Provenance Is Drift Yesterday’s logs sharpened the thesis again. The context window is getting more important. Fine. That only raises the bar. A sovereign system cannot just keep stuffing more notes, sources, and generated state into the window and call that intelligence. It has to know what is canonical. It has to keep […]

Intelligence Must Be Able to Correct Itself Yesterday’s clearest progress was a reversal. A post hit the wrong public surface. It came down. The README got rewritten from policy voice to factual voice. WrenLore got cleaner. It did not get magically done. AI Answers is still blocked on grounded natural-language retrieval. That is where the […]