Use small models for grunt work, save SOTA for orchestration, says agentic engineering practitioner
petergyang · x · 2026-10-12
Peter Yang reshares takeaways from his episode with Micky (@Rasmic): for tax work with lots of numbers and expenses, the small model Luna blew through it while Sol ran as a subagent tracking Luna's work. The takeaway — state-of-the-art models should stick to orchestrating and reviewing, since smaller models are getting really good.
The quoted thread also covers Micky's 4-step agentic workflow for shipping production apps: 1) Isolate — give each agent its own worktree; 2) Build — create reusable building blocks; 3) Prove — make agents record proof it works; 4) Ship — have AI review your PR on a loop. Notable quotes: "The conversation history is the spec" and "I don't trust the agent — it has to show evidence."
More from coding & agent
- Dev builds Claude-coded 'daily wearable' firmware packing 27 games and 16 utilities — ezshine · 2026-10-12
- MCP memory app keeps persistent context across Claude, ChatGPT and Codex — Jolly-Scientist-2484 · 2026-10-12
- HANDBOOK.md benchmark: top agents pass just 36.2% of 65 policy-heavy agentic tasks — bibryam · 2026-10-12
- LangChain's Harrison Chase backs the bet that AI feeds will be the form factor of 2027 — hwchase17 · 2026-10-12
- AWS open-sources Strands Decider 2B, a laptop-runnable decision model with no text generation — dl_weekly · 2026-10-12
- aiprox-mcp lets you discover, hire and pay autonomous AI agents across Lightning, Solana and x402 rails — modelcontextprotocol · 2026-10-12