Agent orchestration: weak model at front desk, strong models in worker threads
dotey · x · 2026-11-17
In a debate over whether the coordinator agent needs a stronger model, dotey distinguishes task-specific agents from bot-app main sessions. He endorses the "weak coordinator, strong workers" pattern: the main session (Sonnet) acts as a front desk that only detects intent, then spins up a Thread forwarding the request to an expert planner (Fable), which in turn delegates concrete execution to worker models (Opus) — like receptionists, a math expert, and grad students. Capability should be spent on planning and execution, not routing.
More from coding & agent
- Baseten launches Voice Agent Stack with sub-second latency at ~$0.01 per audio minute — baseten · 2026-11-17
- rasbt proposes an LLM building-block API to cut switching costs in open-source LLM research — TheTuringPost · 2026-11-17
- Decision Models upgrades playground with image input, showing probabilities and per-request cost — airesearch12 · 2026-11-17
- Microsoft owns every repo on GitHub—so why does Copilot still suck at coding? — prasenx · 2026-11-17
- Baseten open-sources Rust/CUDA DeepSeek engine; agent on Vera Rubin finds optimizations speeding Blackwell by 10% — baseten · 2026-11-17
- shadcn's underrated tip: ask your coding agent "are you sure?" before it writes any code — shadcn · 2026-11-17