TYPESAFE AI launches JEV, its first System One AI model

TYPESAFE AI: TypeSafe AI is an AI company focused on developing models that integrate directly into software for structured decision-making. It is relevant to the news as the organization launching Jev, its first System One model designed to deliver typed outputs with probabilities for use in automation workflows. The company was founded by Diogo Almeida following his work at OpenAI. CompleteSkeptic: Diogo Almeida, operating as @CompleteSkeptic on X, is an AI researcher who co-created RLHF and contributed to the development of ChatGPT at OpenAI. He serves as CEO of TypeSafe AI and is the primary figure behind the two-year stealth development and public release of the Jev model announced in the news. Background: Diogo Almeida previously contributed to foundational AI work at OpenAI before shifting focus to new model architectures at TypeSafe AI. Model Approach: System One models like Jev are trained with a new algorithm called RLCD to produce calibrated decisions that software can directly consume without text parsing. Use Case Focus: The models emphasize parallel evaluation of multiple structured questions against shared state for reliable integration into larger applications.