Percy Liang teaches new MIT course on self-evolving AI agents across four evolution levels
pliang279 · x · 2026-10-23
Stanford's Percy Liang and Ao Qu are teaching a new MIT Media Lab course, "How to AI an AI for Almost Anything," on self-evolving AI systems that recursively improve from experience, feedback, memory and exploration. Core idea: intelligence shouldn't stop improving when training ends.
The course covers four levels of evolution:
- Context evolution — prompt optimization and context management;
- Harness evolution — memory, tools & skills, workflow optimization, multi-agent systems;
- Model evolution — RL, distillation, test-time training;
- (post truncated)
Designed for students from diverse backgrounds, it combines foundational methods, frontier research and hands-on practice building modern agent harnesses, with researchers invited to present each module.
Related event: MIT Media Lab Launches Course on Self-Improving AI Agents(2 posts)→
More from coding & agent
- Grok Bot adds 1Password integration so the agent can sign in for you — belce_dogru · 2026-10-23
- AI agents fuel real-world intrusions: 2,250 M365 tenants hit, 100+ firms compromised — cyb3rops · 2026-10-23
- Playwright's auto-retry hid a UI flash bug — MutationObserver test caught it — SeanOliver · 2026-10-23
- Next.js 16.2 quirk: server action promise resolves before refreshed page data renders — SeanOliver · 2026-10-23
- Four Practical Tricks for Taming Exploding Coding Agent Sessions — srchvrs · 2026-10-23
- LarkAgents launches: first customer saves 50 hours a week automating ERP and reordering — ycombinator · 2026-10-23