MC

Michele Catasta

President & Head of AI

Head of AI at Replit, building the future of software development. Former Head of Applied Research at Google (working on LLMs) and Research Scientist + Instructor at Stanford University.

Writing

2 articles
Diagram comparing a one-off human shipping decision with a continuous evaluation system that learns from production and ships improved agents.

Closing the loop: Evaluating and improving Replit Agent at scale

Most Replit Agent users start with an idea. They describe the goal in natural language — without a repo, test suite, or chosen framework — and expect the agent to turn it into a functioning app. The result might be a website, slide deck, mobile app, several connected artifacts, or something else entirely.