Anthropic Engineer: Multi-Model Pipelines Are a Distraction, 2-3 Models Suffice
brandon_galang · x · 2026-11-15
Anthropic engineer Brandon Galang pushes back on a Googler's elaborate multi-model orchestration workflow, arguing such pipelines are largely a distraction. Most people should just use 2-3 models:
- The best model you can access/afford for core work;
- A Pareto-dominant model for lower-stakes or well-scoped tasks;
- Optionally, a separate model family for diversity in review — though he questions whether that third slot still matters.
The cited workflow uses Fable 5.2 for architecture and ambitious projects, Carbon as the medium-complexity workhorse, GPT-6 Astra/Sol Ultra for layered review and prompt supply, and cheap models for cleanup. Galang concedes other models shine in specific cases (he uses them for personal writing), but thinks the cognitive load of updating your setup with every new release isn't worth it.
A representative practitioner take: the returns from model orchestration may be overrated; a simple, stable two-model setup is more practical.
More from coding & agent
- AI meme hits home: under all the frameworks, it all boils down to one SQL statement — teodorio · 2026-11-15
- Aaron Levie: work will become coding-shaped to fit what AI agents are good at — RaphaelDabadie · 2026-11-15
- msgblast ships iPhone app with local MCP, agent messaging via Tailscale, and Amazon shopping agent — msg · 2026-11-15
- Building replayable, resumable workflows for long-running AI agents: lessons from Pyash — aizvo · 2026-11-15
- Permission Delta: Open-Source GitHub Action Reviews MCP Config Changes in PRs — AGENTPROOFINC · 2026-11-15
- How Do You Excavate Extremely Long ChatGPT Conversation Archives Without Losing Chronology? — clearbreeze · 2026-11-15