Now
- Training a small reasoning model, and building the tools to watch it think.
- Building an LLM inference engine from first principles, and turning the build into a masterclass on how serving actually works.
- Taking coding-agent harnesses apart down to their primitives: sessions, tools, memory, sub-agents.
- Research on how AI systems get judged, and where the judgments break.
- Long-form system builds and daily videos on YouTube.