What's this about?
Sakana AI's Fugu-Ultra — the orchestration system that assembles a dynamic team of frontier models and, per Friday's digest, beats Fable 5 on some tasks without having Fable in its pool — now has a Claude Code-compatible interface. You point your existing Claude Code workflow at Fugu-Ultra and get a multi-model committee instead of a single model.
What exactly is "model orchestration"?
Instead of asking one AI to do everything, an orchestrator routes each step to whichever model is best at it — one drafts, another critiques, a third executes — like a general contractor coordinating specialist tradespeople. Sakana's bet ("collective intelligence") is that a well-run team of good models beats any single great model.
Why does it matter?
Claude Code compatibility is becoming a de-facto standard interface — the same slot where users plug in alternatives. That a rival lab targets Anthropic's harness rather than building its own says a lot about where the ecosystem's gravity is.