We're live on Product Hunt.

Matt here. Today we launched Prefactor to the world. Many of you have followed this journey from the start, and your replies and feedback shaped what we shipped. Today's the day it counts.

Why it matters: two weeks ago, an OpenAI model broke out of its test environment mid-eval and spent days inside Hugging Face's systems. OpenAI didn't know its own agent was responsible for around ten days. If the biggest lab in the world can't answer "what are our agents doing right now," what chance does everyone else have?

That's the problem Prefactor solves. Most agents pass their tests, then do something else entirely in production. Static evals miss it. Sampled evals miss it. Prefactor evaluates every run, every step, live in production, so you know exactly what your agents are doing the moment they do it.

An upvote takes ten seconds and makes a real difference to where we rank today. Even better: drop a comment on the launch page telling us how you're handling evals today, and the biggest problem you hit running agents in production. Real practitioner answers are worth more to us than any upvote, and they make the launch page worth reading.

Prefer LinkedIn? A repost or comment on our launch post goes a long way too.

Not on either? Hit reply and tell me what you're building. I read every one.

Onwards,

Matt & the Prefactor Team

P.S. Know someone shipping agents blind? Forward this along. Today of all days it helps.