finally vibe motion design
By my tests, not a one model (Astra, Grok 4.7, Kimi K3, Deepseek V4.1 Flash) from other companies could beat Opus 5.5 in creating motion design.
But, Astra and Gemini 3.8 flash beat Opus 5.5 in sound design without context for a finished motion video
i guess i built the fastest inference engine (ftt-rs) for fasttext models
~6.1μs prediction time per text (~2.5x faster than fasttext-turbo)
~4.7ms to load lid176 model (~40х faster than fasttext-turbo and ~400x faster than fasttext)
written from scratch in Rust
A team of 1,000 Fable 5.2 agents fully formalized Fermat’s Last Theorem in 30 hours, translating the proof into a language a computer can check step by step.
With the same setup, Opus 5.5 completed only 26% in 18 hours, and barely made any more progress after 30 hours.