Simon WillisonWednesday · September 23, 2026FREE

llm-typesafe 0.1a0

Simon Willison published llm-typesafe 0.1a0, described as an LLM plugin for accessing Jev and other TypeSafe AI models. The release adds support for TypeSafe AI's new Jev model to the LLM tool. Installation is done with "llm install llm-typesafe", after which users set an API key with "llm keys set typesafe"; Willison notes the waitlist for obtaining a key "seems to move pretty fast." The post shows three example invocations. A yes/no "noul" question runs "llm -m jev" with a system prompt asking whether a message explicitly requests a refund, returning output of the form {"type": "noul", "noul": 0.99}. A choice question pipes a message file into the model with a system prompt about which team should handle it, using "-o answer_type choice" and a criteria object mapping billing, technical, and other categories. A scoring question pipes a report file with a system prompt about reproducibility, using "-o answer_type score" and a criteria list ranging from no reproduction instructions to complete steps with expected and actual results. Willison directs readers to the README for more details. The post is tagged with projects, llm, and jev, and appears alongside his recent articles on Claude Opus 5.5, GPT-6 Sol, GPT-6 Luna, a new price war, and Jev introducing a new shape of LLM.