⚡ Jev is HERE and it's rejecting ChatGPT's entire training playbook ♠ built by a co-creator of RLHF/ChatGPT who says

Evaluationby fahdmirza2026-09-15

⚡ Jev is HERE and it's rejecting ChatGPT's entire training playbook ♠ built by a co-creator of RLHF/ChatGPT who says chat was the wrong bet 🔹New training method: RLCD (Reinforcement Learning for Calibrated Decisions) instead of RLHF 🔹Not chat — typed decisions with calibrated confidence scores, built for software not humans 🔹Claims: 193x faster, 444x cheaper, 0% structured-output error rate vs GPT & Claude 🔹$42 per billion input tokens — output tokens free 🔥the full breakdown below:👇 https://youtu.be/X8Outd-khS0?si=VF6cv4HOj5r3V_6-

Links

Jev: The Model That Killed Chat GPT's Core Idea? RLCD Explainedyoutu.beThis video introduces new model Jev, based on RLCD — Reinforcement Learning for Calibrated Decisions.📬Weekly AI Newsletter: https://fahdmirza.substack.com/?...

Jev: The Model That Killed Chat GPT's Core Idea? RLCD Explainedyoutube.comThis video introduces new model Jev, based on RLCD — Reinforcement Learning for Calibrated Decisions.📬Weekly AI Newsletter: https://fahdmirza.substack.com/?...

open source discussion

More in Evaluation

Return probabilities, not labels — 80/10/10 beats "orange"I’m building Aurapunk, an open-source multi-agent IDE.Tested jev against deepseek v4.1 flash on norwegian text.