Skip to content

AI education

Agentic applications

One system you can leave running. Memory, tools, a sandbox, a schedule, evaluation, and a hard line on what it may do.

19 parts · ~102 h

Start

Lessons

  1. 00Course 2: agentic applications on local models — index
  2. 01Part 1 · Agent architectures — a map of the terrain
  3. 02Part 2 · The skeleton — process, configuration, sessions
  4. 03Part 3 · Harness — a loop that survives long tasks
  5. 04Part 4 · Context engineering — the attention budget
  6. 05Part 5 · Long-term memory — three schools
  7. 06Part 6 · Identity — working files as the program
  8. 07Part 7 · Tools — design and ergonomics
  9. 08Part 8 · Code-mode — the agent writes a program instead of calling tools
  10. 09Part 9 · Protocols and integrations — MCP, CLI, A2A
  11. 10Part 10 · The production sandbox
  12. 11Part 11 · Scheduling — an agent that wakes itself up
  13. 12Part 12 · Channels — how you talk to the agent
  14. 13Part 13 · The fleet — many agents, one system
  15. 14Part 14 · Evaluation — how you know this works
  16. 15Part 15 · Observability — seeing what it does
  17. 16Part 16 · Agentic security — OWASP ASI01–ASI10
  18. 17Part 17 · Hybrid and router — local by default, cloud by conscious choice
  19. 18Part 18 · Release and maintenance
  20. 19Part 19 · Capstone — a general-purpose personal assistant
  21. —Sources — what this course stands on