Jev, a classification-only model from TypeSafe, scales a news workflow from 20 to 500+ items

every · x · 2026-09-28

TypeSafe, an AI lab in stealth for two years, released Jev on September 15—a model that classifies text and structured data instead of generating it. The pitch: it's fast, cheap, and accurate enough to handle judgment calls you'd never pay an LLM to make at scale (is this email urgent, does this sound AI-written, should this ticket be escalated). X demos have gone viral: playing games, re-ranking search, filtering feeds, plus a 1M-view demo inferring intent from speech and gestures.

Every's marketing head Douglas Brundage found a post showing Jev scanning news for brand mentions, handed it to his coding agent with "can you just do this with Jev?", and scaled his morning news review from under 20 posts per run to 500+, including sources beyond X.

The reusable shortcut: find a demo close to what you need, give it to your coding agent, and ask for a testable version.

Original post →

More from coding & agent

coding & agent channel →