Reading list
Links
Things I've found worth keeping — articles, repositories, posts, papers. Captured day by day from my journal, then curated.
18 of 18
Notes on AI Apps in 2026 (a16z)
The building-layer thesis, 'thick apps', and the shift from execution to ideation. The app layer thrives even where the labs focus.
The Third Era of Software Development (Cursor)
Michael Truell on agent fleets working autonomously; usage inverted from tab-completion to autonomous agents.
Effective context engineering for AI agents (Anthropic)
Context as a finite resource with diminishing returns; progressive disclosure and just-in-time retrieval. The discipline that replaced prompt engineering.
Equipping agents with Agent Skills (Anthropic)
The thin-harness, fat-skills pattern made concrete: a SKILL.md folder loaded by progressive disclosure.
Effective harnesses for long-running agents (Anthropic)
Stateless sessions recover work by persisting state to git and progress files: initializer agent plus coding agent.
Horseless Carriages (Pete Koomen)
Why most AI apps fail: they bolt AI onto an unchanged product instead of inverting the architecture around the model.