Skip to content

Add agent-building insights from Addy Osmani's 2026 blog posts - #7

Open
lsimons wants to merge 1 commit into
addyosmani:mainfrom
lsimons:contrib/agent-building-insights-2026
Open

Add agent-building insights from Addy Osmani's 2026 blog posts#7
lsimons wants to merge 1 commit into
addyosmani:mainfrom
lsimons:contrib/agent-building-insights-2026

Conversation

@lsimons

@lsimons lsimons commented Jul 19, 2026

Copy link
Copy Markdown

Initial AI prompt: Scan Addy's blog posts from this year at https://addyosmani.com/blog/ . See if any of them are explicitly about building agents. If so, read them, and see if they have any great advice or content that's missing from this course, and add it.

Resulting additions:

  • Lesson 7: the orchestration tax - human review capacity as the ceiling on multi-agent parallelism
  • Lesson 9: mistake 7 - never let the agent grade its own homework; review test diffs more carefully than code diffs
  • Lesson 10: calibrated autonomy - three risk questions and a pre-delegation contract
  • Lesson 13: the harness concept (agent = model + harness) and the ratchet pattern
  • Lesson 15: write AGENTS.md by hand - the discovery filter
  • Lesson 17: workflows not essays, anti-rationalization tables, evidence as exit criterion
  • Lesson 18: long-running agents - the three walls, externalized state, and the Ralph loop

Each addition links the source post in further reading.

Ported from lsimons/agent-engineer-course commit c3d862f .

Assisted-by: Claude:claude-fable-5

Ported from lsimons/agent-engineer-course commit c3d862f, adapted to
this repo's layout (NN-name/README.md, relative links) and ADK-based
lesson content:

- Lesson 7: the orchestration tax - human review capacity as the
  ceiling on multi-agent parallelism
- Lesson 9: mistake 7 - never let the agent grade its own homework;
  review test diffs more carefully than code diffs
- Lesson 10: calibrated autonomy - three risk questions and a
  pre-delegation contract
- Lesson 13: the harness concept (agent = model + harness) and the
  ratchet pattern
- Lesson 15: write AGENTS.md by hand - the discovery filter
- Lesson 17: workflows not essays, anti-rationalization tables,
  evidence as exit criterion
- Lesson 18: long-running agents - the three walls, externalized
  state, and the Ralph loop

Each addition links the source post in further reading.

Assisted-by: Claude:claude-fable-5
Co-Authored-By: lsimons-bot <bot@leosimons.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant