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
Kaustav Banerjee shares a complete automated workflow for an access ticket: 4 of 20 steps involve judgment calls handled by TypeSafe AI's Jev, not an LLM. The LLM runs once, the whole ticket costs ₹0.16, and Jev's share is under 1 paisa.
The author upgraded AI moderation triage in Bitsocial communities from gpt-5.6-luna to gpt-6-luna, with 50% cheaper input, 58% cheaper output, yet smarter. Jev still handles easy checks, and uncertain ones go to grok-4.7.
The founder of Jev released a 1-hour masterclass breaking down how Jev actually works: LLMs → Decisions → Verification → Coding Agents, covering why LLMs alone aren't enough, the three primitives behind Jev, and turning huge AI tasks into tiny reliable steps.
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.
A user shares a slow-motion clip of Jev solving a Sudoku puzzle.
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.
The community shared a way to run Jev on Cloud Run with an RTX PRO 6000: ~47s cold start, ~117ms end-to-end, and 100+ req/s at concurrency 32. They also used Jev to build Snake, Dino, and Tetris games.
JEV is being integrated into Ekko Agent's memory system, participating in key stages such as semantic classification, candidate memory re-ranking, and pre-write validation. It helps the agent find more relevant memories and be more cautious about what to remember. Currently in testing, early feedback shows improved recall accuracy in some scenarios.
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 combines the recently released Claude Opus 5.5 and Jev for a marketing use case: automatically scoring 120 competitor ads, shortlisting 10 worth adapting, and creating versions of each.