Ex-Amazon engineer's 10-principle manifesto: ship faster by rebuilding your workflow for agents
blaizedsouza · x · 2026-09-11
Clare Liguori, an engineer who worked with AI coding teams across Amazon, published a manifesto arguing that developers have split into two groups: those who merely swapped coding tools, and those who changed how they work — and only the latter see step-function productivity gains.
Core thesis: you are no longer building the software directly; you're building the agent setup that builds the software. Her ten principles include: be the architect, not the typist; maximize agent time and minimize your involvement; build your codebase for agents; give agents a fast feedback loop; execution is cheap, direction is everything; treat code as disposable; hold AI output to human standards; trust the boundaries, not the agent; use agents for everything, not just code; and continuously tune your agent setup.
She stresses this is not vibe coding but a disciplined engineering investment — expect the first weeks to feel slower as you write steering files, refactor for agents, and learn task decomposition, before it gets dramatically faster. The lasting skill, she argues, is managing your own attention across multiple parallel agents. Published as part of Kiro's Frontier Engineering guide.
More from coding & agent
- NVIDIA and Red Hat host 33B-cap Small Models Hackathon in Raleigh, vLLM adds a bounty — vllm_project · 2026-09-11
- Building an AI news radar with Doubao: auto-scrape hot items into a Feishu group — songguoxiansen · 2026-09-11
- Nebius launches AI Builder Program: free agent courses, $1 NVIDIA certification, $400+ credits — heyshrutimishra · 2026-09-11
- Early MCP contributor who shipped it at Clerk now strongly advises against the protocol — cnakazawa · 2026-09-11
- Cricket-for-noobs MCP server: 15 tools, win probability from 8,000+ matches, baseball-style explanations — Turbulent_Ice_8710 · 2026-09-11
- ChatGPT Desktop Codex app can now use local models via Ollama 0.34 — ollama · 2026-09-11