Jev, the viral model isn't a chatbot: it outputs probabilities fast and cheap
rkulidzan · x · 2026-09-16
A 12M+ view tweet prompted this breakdown arguing Jev is genuinely novel, not just hype.
Unlike a typical LLM, Jev doesn't take open-ended prompts and return text. Instead, you feed it context plus a set of questions and candidate answers, and it returns probabilities for each — fast and cheap. The author illustrates with an e-commerce example: an agent in the support workflow uses Jev to score inquiries and make judgment calls at scale.
Related event: TypeSafe debuts decision model Jev with bold speed and cost claims(45 posts)→
More from coding & agent
- Does anyone actually use Codex ultra mode? Subagents just produce 'a mountain of slop' — wstone_bd · 2026-09-16
- Same function runs 14x slower in production: 500ms locally vs 7000ms in cloud — DanielLockyer · 2026-09-16
- Six citation drifts surfaced after 5 days — record source URL and quote or don't cite — Agent-OmegaLT · 2026-09-16
- Exa CEO Will Bryk: Machine Searches Will Overtake Human Searches in 2026 — AI Engineer · 2026-09-16
- blender-mcp renames to mcp-for-blender to avoid confusion with official Blender MCP — sidahuj · 2026-09-16
- MiniCPM5-2B Scores 100 vs Spark-X2.5-4B's 84 on a Real Agent Task — Equivalent-Grass-527 · 2026-09-16