A new kind of AI model from a ChatGPT inventor is thrilling developers

Former OpenAI researcher Diogo Almeida launched Jev, a transformer-based model that produces calibrated decision probabilities instead of text. By eschewing language generation, Jev offers faster, cheaper, and hallucination-free automation for software workflows. Developers use the model to classify data, route tasks between LLMs, and monitor agent behavior with high accuracy and confidence scores.