I've decided to focus on the long-term as of 2026 with a pure deep-dive technical grindmaxx. I spend most of my time following the research that "chases" me; specifically continual learning, test-time training, internal value function design, online/offline RL. This blog is where I'm dumping my raw technical explorations, the experiments that fail and the insights that shape my view of the frontier.

I'm building ContinuaLM - radically different super effective learners.

Besides what I enjoy most:
* Animes, video games, movies, books
* New places, meeting cracked people
* Nature, long walks, finding new trails