Blog
Notes from the model lab.
Engineering writing on latency, evaluation, privacy, and running AI in production. The first posts are on the way.
How we benchmark a model before it ships
Coming soonCutting p99 in half without touching accuracy
Coming soonIrreversible face templates, explained
Coming soonServing five models behind one API
Coming soon
Want these in your inbox?
We’ll send new posts when they’re published — no noise, just the engineering.
Get notified