Halluminate is building reinforcement learning environments and benchmarks to help AI models do knowledge work beyond coding, starting with finance. With a team of fewer than 10 people, the company works with four of the top five closed-source US AI labs and recently raised a $30M Series A. In... this fireside, co-founders Jerry Wu and wyatt marshall sit down with YC's Jon Xu to share how testing browser agents led them to building the environments models now use to practice tasks like financial modeling. They explain why better verification, not just more training tasks, is critical to making models smarter, and how that work could eventually lead to simulated companies where teams of AI agents learn to work together. 00:00 — Halluminate’s Pivot From Evals to Training Environments 03:48 — Teaching AI to Do Financial Work 07:13 — Why Verification Matters More Than Volume 10:13 — From Finance to Simulated Companies 13:30 — AI Safety and Building the Teamshow more

finance is where every model discovers its spreadsheet is hallucinating.

The important leap is not just training models on harder domains. It is giving them environments where decisions can be tested, mistakes become data, and progress is measurable. That is how AI moves from impressive demos to dependable work.

verification before scale ✅

AI models handle finance data well too. RL environments help them learn complex rules quickly without much human effort. It works for me too.

Thanks for having us! We wouldn't be here without the whole @ycombinator team.

The pivot from evals to training environments is a good example for founders. They tested browser agents, saw what was missing, and then built that missing thing. Good ideas often come from the work you were already doing.

Less than 10 people, four of the top five labs, $30M in. They're selling shovels and the gold rush just started.

Finance is a smart first market because a model either ties out or it doesn't. Marketing will be the hard one. Half the job there is agreeing on what "worked" even means.