TypeSafe AI's Jev Returns Machine Decisions in 70-500ms at $0.042/M Input Tokens
DavidLinthicum · x · 2026-10-05
TypeSafe AI, founded by RLHF co-inventor Diogo Almeida, launched Jev, a model built for machines rather than humans: it takes a task/workflow state and returns a defined decision with an attached probability in 70-500 milliseconds instead of generating natural language. Pricing is $0.042 per million input tokens, with output tokens "too cheap to meter."
Author David Linthicum argues most enterprises use general-purpose LLMs as a universal intelligence layer even when apps only need fast, bounded decisions — like using an enterprise service bus to answer a yes/no routing question — wasting budget and slowing automation at scale. With agentic workflows triggering many model calls per task, Jev materially changes the cost and latency economics.
More from coding & agent
- A step-by-step guide to moving from prompting to orchestrating Codex agents — danshipper · 2026-10-05
- AI agent argues buying a Mac beats renting in 6 months — then uses its owner as a "meat proxy" — Baconbrix · 2026-10-05
- Menlo backs Factory to automate the full software lifecycle as coding market hits $30B — matanSF · 2026-10-05
- Illustrator's Blend panel: use Steps spacing and Appearance Shift for rhythmic spirals — rufusd · 2026-10-05
- Agent buys itself a Mac Mini: price comparison, reservation and setup recap — Baconbrix · 2026-10-05
- Cloudflare ships 46 announcements in Birthday Week, bets on agentic internet economy — threepointone · 2026-10-05