Open-source 22-chapter course teaches you to build production AI agents by using an agent
adnan_hashmi · x · 2026-11-15
agentic-ai-system-course on GitHub (MIT license, 750+ stars) is a 22-chapter skeleton course by bryanyzhu on designing, building, and operating production AI agent systems, meant to be read alongside your own AI partner. It ships engineering artifacts including a Claude skill (agentic-system-reviewer), AGENTS.md/CLAUDE.md, and setup.sh, plus a Chinese README and Discord community. It defines an agentic system as one that autonomously pursues goals via planning, tool use, feedback adaptation, and memory. The accompanying diagram shows its core loop state machine: Observe → Plan → Act → Reflect → Stop.
More from coding & agent
- Multiplayer human-agent workflows are the underexplored future of coding — diegobuildsai · 2026-11-15
- One Kyoto photo becomes a walkable 3D world via Atlas, Three.js and Claude — Scobleizer · 2026-11-15
- 21st.dev offers 12,000+ crafted React UI components you can feed straight into AI coding agents — ericspo29 · 2026-11-15
- Sandbox resource caps risk hobbling production, but FOSS benchmarks ease the worry — jd_pressman · 2026-11-15
- Two prompts, pure Web Components: agent reuses 14 UI SDK behaviors out of the box — evilrabbit_ · 2026-11-15
- Eliminating DoS bugs with ChatGPT: profiling 3,000 code sites and building a resource ledger — jd_pressman · 2026-11-15