What happened
Simon Willison released llm-typesafe 0.1a0, an LLM plugin adding support for TypeSafe AI's Jev model, installed with "llm install llm-typesafe" and an API key set via "llm keys set typesafe".
Why it matters
The plugin gives LLM users three question formats for Jev — yes/no "noul" questions, choice questions, and scoring questions — so messages and reports can be routed or rated through the tool.
What to watch
The model is behind a waitlist, so access hinges on getting a TypeSafe API key; the README is the place to check for setup details.
WHO IT HITSDevelopers and technical teams already using the LLM command-line tool can now send messages and reports to TypeSafe's Jev model for yes/no, choice, or scoring answers, provided they get an API key. Non-technical staff are unlikely to touch this directly, since it requires the LLM tool and key setup.
Summaries like this, in your inbox every morning.
Simon Willison built llm-typesafe as a new plugin for his LLM tool, specifically to add support for TypeSafe AI's Jev model. Installation is a single command, and setup requires an API key stored through the LLM tool's key command. The post points readers to a waitlist for that key.
The plugin exposes three question types through the LLM command-line interface. A yes/no "noul" question returns a JSON object with a noul value, choice questions take a criteria object mapping labels to descriptions, and scoring questions take a criteria list of levels. These are shown with example prompts about refunds, message routing between billing and technical teams, and judging how reproducible a problem report is.
What the outcome hinges on is access: the model sits behind a waitlist, so whether this gets used depends on how quickly keys are handed out. For people already working inside the LLM tool, the plugin is a small, low-friction way to try a different model for classification-style tasks. Details beyond the post itself, such as pricing or limits, are not stated here.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, LINE, or Slack.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
DeepSeek CEO Liang Wenfeng told investors the startup aims to rely on Huawei chips to train its models
Arm estimates agentic AI could raise data-center CPU demand from roughly 30 million cores to about 120 million…
Shopify is partnering with Meta so Muse, Meta's AI shopping agent, can complete purchases at Shopify merchants…
NVIDIA released Isaac ROS 5.0, adding agentic workflows, Isaac Skills for setup and manipulation, and support…
Anthropic launched Claude Opus 5.5, saying it matches Claude Fable 5.1 on most tasks while costing about 40 pe…
Anthropic debuted Claude Opus 5.5, which it says matches flagship Fable 5.1 performance at around 40% less cos…
