TypeSafe AI launches Jev for software automation

TypeSafe AI has launched Jev, a specialized artificial intelligence model designed for software automation and classification rather than natural language generation. Founded by former OpenAI researcher Diogo Almeida, the startup developed Jev to address the limitations of Large Language Models (LLMs) in machine-to-machine workflows.

Unlike traditional LLMs that generate text, Jev functions as a classifier. It accepts an input and a predefined output schema, returning “calibrated decisions” in the form of probability scores. This architecture is intended to eliminate AI hallucinations by providing structured, numerical outputs that software can act upon directly.

Early testing indicates significant efficiency gains. Developers have reported that Jev can process tasks, such as email triage, at a fraction of the cost of traditional models—potentially up to 100 times cheaper. In some benchmarks, the model has demonstrated speeds 5 to 18 times faster than existing models while maintaining high accuracy for structured decision-making tasks.