Sakana AI has released Fugu, a multi-agent orchestration system that breaks user prompts into subtasks and routes them across a swappable pool of expert models via a single OpenAI-compatible API. Sakana claims Fugu matches frontier models like Anthropic's Fable 5 on benchmarks and positions it as a blueprint for AI sovereignty — reducing reliance on any single provider. However, early community reactions are skeptical: users report slow API speeds, fast token burn rates, high pricing ($5/$30 per million input/output tokens for Fugu Ultra), and real-world performance that doesn't always match benchmark claims. Critics argue that since Fugu still depends on third-party models, it can't truly deliver AI sovereignty if multiple providers restrict access simultaneously. The debate touches on whether 'collective intelligence' via orchestration is the future of AI, or just an expensive black box layered on top of other black boxes.
