Three tracks on building, distributing, and operating agent systems — written for engineers shipping this stuff at work, not for people watching keynotes.
Track 01 is the foundations: transformers, RAG, tools, sandboxes, the production loop. Track 02 is what most agent material skips — multi-agent systems are distributed systems, and there are fifty years of theory that explain exactly how they break. Track 03 is the operating-system layer underneath: scheduling, isolation, resource control for fleets of agents.
3 tracks · 17 modules · 10 lessons published · free · MIT-licensed · no signup
The same problem from three angles. You can read in order or skip straight to whichever one matches what you’re fighting today.
From transformer foundations to production agent operations.
Your agent already is a distributed system. Here is the theory that explains why it breaks.
Scheduling, isolation, and resource control for fleets of agents.
No course platform. No paywall. Code that runs. Opinions where the field is messy.
No signup, no email gate to read, nothing to buy. The repository for Track 01 is open on GitHub at the same SHA the site renders. Fork it, run it, teach a course off it.
Every code example that ships to a lesson is validated end-to-end in research/code-validation/ — Python via uv, TypeScript via tsc — before it lands. Hallucinated APIs that look plausible are the failure mode I’m most paranoid about.
Track 01 is the foundation: build agents that hold up. Track 02 is what most material skips: multi-agent systems are distributed systems, and the theory exists. Track 03 is the operating-system layer underneath, scheduled for later. You can read in any order.