[AINews] Jev: a “System One Model” that only decides/classifies/routes/scores — >100x faster, >200x cheaper than small frontier LLMs

AIEi Paris (Sep 23-24) and AIE NYC (Oct 12-14) is >50% sold out, AIE CODE (Nov 10-12 in SF) and AIEi Shanghai (Nov 5-6) are next on deck before AIEi Sydney (Dec 7-8 alongside NeurIPS) closes the year! It’s very rare that a new startup launch will make title story, especially on a day when Gemini 3.8 Live and Periodic Labs had strong announcements, however, TypeSafe’s launch has sat comfortably atop Hacker News all day. We were fortunate to preview them last month at AIE pre launch: After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x … 6:17 PM · Sep 15, 2026 · 4.21M Views For those used to traditional autoregressive LLMs, a fast model that cannot code and doesn’t reason might feel counterintuitive in its usefulness. That’s exactly what the team is aiming for in complementing “System Two” slower LLMs: you let go of strings and chat, and you get 1) parallel sampling, 2) “no hallucination”,…