Free Lesson
Build a PM Skill that Learns from you (Learning Loops)
45 min
Apr 24, 2026 12:00 PM
What you'll learn
If agent loops can handle context, what must PMs design
Understand learning loops that are still an open area of research and are critical for building scalable AI agents.
Hands-on: Build agents using Claude Code
See how even the latest models (Opus 4.7) struggle with tasks that require continuous, human-level improvement
Explore the risks of deploying agents without learning loops
Learn the system design patterns that address poor UX
Learn how to design feedback-driven learning loops
Apply learning loop system design via a hands-on case study you can use in real products and interviews.
Why this topic matters
For 70+ years, programming meant humans writing code & machines executing it. Autoresearch, Ralph, & Hermes invert this: humans write natural-language instructions (program.md, prd.json, SKILL.md), & the agent writes, modifies, and iterates on code. This isn't vibe coding — it's a new job description. The PM becomes a "programmer of programs.md" — authoring research org instructions, not Python.
You'll learn from
Previously at