In one of my classes, I asked, “Who is afraid of not getting a job after graduating?” About 80% of the 150 students raised their hands. PhD students email me wanting to leave academia for a frontier lab, believing research there is meaningless. I believe both are wrong. They assume the future belongs to whoever has the most GPUs. I think academia is about to have a renaissance. This week is our argument in code.

With agents, research projects have become easy and quick. When every project is easy, piecemeal work stops being good research. The unit of research is now the ecosystem, not the paper. That is what Open Source Week is for. My lab uses academic freedom to bring the best AI tools to everyone for free.

We built open systems, made models cheaper to run locally, and created new methods for domain-specific RL environments. All of it sits at the intersection of inference-serving frameworks, agent harnesses, and autonomous research systems. Accessibility has two halves — resources and expertise — and only one is fixed by hardware. A couple of GPUs or a MacBook can be enough.

We will show three things: frontier autonomous research, the most efficient test-time scaling I know of, and auto-compaction far more efficient than Claude Code or Codex.