THE JEV GUIDE · CURATED EDITION
Launch news, explanations, demos, and community projects around TypeSafe Jev, organized by topic. Every entry links back to its X source.
Curated selection · Not exhaustive · Updated manually
01 / EXPLORE
0 means awaiting a verified post
Based on @karminski3's test framework, the author adds multiple maze test results for Jev and compares them against random walk and professional maze algorithms.
The author used TypeSafe's new model Jev in Skills IL to build a tool that detects bank phishing messages received via SMS, WhatsApp, or email. Users can paste a suspicious message directly to verify it.
Someone built a massive directory of everything related to Jev, including skills, repos, X/Reddit posts, resources, sites, and more, filterable by use case.
The author suggests that Jev's return values should not be treated as simply correct; instead, focus on confidence. For Choice types, a cannot_tell option should be added to avoid misjudgment.
The author thinks Jev is not suitable for AI-generated content detection, all AI detectors are similar, and points out it's pointless unless accuracy exceeds 90%, recommending another API. Link
The author tried having Jev solve a Sudoku, noting the prompt was tricky but the speed was fast, and mentioned LLMs are said to struggle with Sudoku.
The author says Jev 1.13 might replace large language models for some tasks.
Rajeshsam notes that AI momentum seems to have circled back to traditional data science with System 1 classifiers (e.g., Jev). He argues that typed structured responses plus deterministic reasoning in Python via hooks beat letting a model guess control flow; leave fuzziness at the ingestion boundary.
A community rerun of hard fact-check benchmarks finds Jev and GPT-5.4 statistically indistinguishable (3.3-point gap), while Jev costs about a fiftieth as much. Link
Developer Tal Saiag shares a Jev pattern: each call deliberately leaves a small piece of state only for the next Jev call, which he calls 'Jev carry state'.
A user shares learnings from an experiment with Jev: it requires explicit prompting like older GPT-4-level models, is very cheap, and they tuned it for about 8 hours with Astra.
Wildan tweets that it's time for Claude to use TypeSafe AI's Jev.