Jev: System One models for Prod, not God — with Diogo Almeida, CEO, TypeSafe AI
Diogo Almeida launches Jev, a new model paradigm from TypeSafe AI challenging autoregressive chat-tuned LLMs.
“we have dropped every mode other than autoregressive chat-tuned LLMs because of the overwhelming success of ChatGPT”
InstructGPT coauthor Diogo Almeida, now CEO of TypeSafe AI, launched Jev, a non-autoregressive 'System One' model paradigm positioned for production use and aimed at fixing alignment, refusal, and reliability problems he attributes to chat-tuned LLMs. The launch video drew ~40M views and dominated the AI timeline, with emerging use cases in computer use, gaming, coding agents, and analytics. It matters as a credible frontier-researcher bet against the dominant autoregressive chat LLM approach.