Ryan Tuffy

[Ryan Tuffy @TheRyanTuffy]

@DavidVaugh38865 Thanks @DavidVaugh38865. Send me a DM if you end up trialing this in your stack. I'd love to hear how it worked out for you!

[Ryan Tuffy @TheRyanTuffy]

Most LLM calls in production aren't really generation. They're decisions: route this ticket, flag this prompt, score this urgency.
Laya handles those in one forward pass: ~33ms, calibrated probabilities, zero output tokens.
Open weights, Apache 2.0.

https://t.co/yJmHwyE5Un

[Ryan Tuffy @TheRyanTuffy]

Paper was published in Sep 2025 as well :)