Ex-ChatGPT co-inventor launches Jev, a decision model claiming 20-200x speedup over LLMs
rohanpaul_ai · x · 2026-09-17
An author claiming to have co-invented ChatGPT and spent two years in stealth on a new training method (RLCD) announces Jev, a new model type optimized for making software decisions rather than generating text.
- Claimed 20-200x faster and 40-400x cheaper than LLMs (with free output tokens) by skipping token-by-token generation
- Instead of generating text that software must parse and validate, Jev takes data plus a predefined question and returns a typed choice or score with probabilities and confidence
- Described as a smart if-statement: a support app can ask if a ticket is urgent, policy-violating, and which queue it belongs to, then act directly
- All figures are vendor claims, unverified by third parties.
More from Models
- 'Grok 4.7 is in the air': xAI community senses a new model release imminent — mark_k · 2026-09-17
- Typesafe AI's Jev saturates classifier eval and runs 6x faster than Gemini 2.5 Flash Lite — cramforce · 2026-09-17
- Netizens speculate on rumored GPT-6 Sol as a lightweight 'mini-Astra' for agentic AI — mark_k · 2026-09-17
- ChatGPT users frustrated as message editing disappears, forcing new chats — Innomen · 2026-09-17
- Code Arena to Reveal Win Rate of #1 GPT-6 Astra vs #2 Claude Fable 5.1 — arena · 2026-09-17
- Stealth Model Union Alpha Free for a Week, Built for Agentic Coding — AiBreakfast · 2026-09-17