Echo: Fable-level results from open models, says Show HN

What's this about?

A Show HN claims to match Fable 5 quality at a third of the cost by pooling open-weight models (GLM-5.2, Kimi, others) behind one interface — the same ensemble bet as Fugu, from a two-person startup.

What exactly is the claim's basis?

Ensembling: different models fail differently, so routing or combining them can beat any single member. OpenRouter published similar findings recently. The catch is showing it holds on *your* tasks — which is why the thread demands benchmarks.

What's the community saying?

1. Launch-quality roast: no benchmarks, no model list, AI-generated demo video, credit card required — "premature" is the polite version.
2. "Time is a flat circle": greybeards recall Dogpile, the meta-search engine over AltaVista and Lycos.
3. The economics objection: Claude subscribers get ~$2500 of Fable usage for $200/month, so "1/3 of API price" mis-targets who actually pays.
4. Someone checked the "backed by YCombinator" claim and couldn't find them in the directory.