The Jev launch post
Diogo Almeida introduces Jev, a structured decision model from TypeSafe. This resource provides an overview of the model's core approach, practical examples, and performance metrics as reported by the author.
The best of Jev. Real projects, practical guides and ideas from across the internet.
Diogo Almeida introduces Jev, a structured decision model from TypeSafe. This resource provides an overview of the model's core approach, practical examples, and performance metrics as reported by the author.
TypeSafe officially launches its research lab and introduces Jev to the developer community. This announcement features a launch film by Diogo Almeida and provides foundational context for the project's mission and future direction.
TypeSafe has removed the waitlist for Jev, allowing public access to the platform. Developers can use the official console to integrate and test typed decision-making capabilities within their own software applications.
TypeSafe highlights Jev’s typed-output approach, where applications provide state and questions to receive structured probabilities, choices, or scores. The announcement also details the integration of the Venice API for developers.
Start with the Launch Post. Explore four official signals.
AI-assisted summaries and translations. Check original sources for context and performance claims.
This Brave browser extension analyzes social media posts to identify logical fallacies in real time. It marks content with a green flag when it detects well-reasoned arguments, aiming to assist users in evaluating the quality of online discourse.
Async LangGraph workflow that gets a typed Jev Choice (invoice or general) and routes each inbound email to the matching handler.
This demonstration showcases a workflow where Jev dynamically routes tasks to different AI models. It illustrates how users can streamline interactions by selecting specialized models directly within a single terminal interface.

Jev plays Puyo Puyo. A community post describing an experiment where Jev makes decisions in a game by evaluating board states and strategies.
This project demonstrates an ambient assistant built with Jev that functions without a wake word. It utilizes probability-based analysis to identify user intent within continuous audio streams, aiming to create a more natural interaction model.
A user creates music by giving real-time voice commands to Jev.
A challenge where 100 builders used Jev and Fable 5.1 to create a post scheduler, highlighting Jev's speed.

Guideless is an AI-powered platform that turns software workflows into video training guides.
Explore sponsorBounded exploratory browser testing with Jev, deterministic assertions, and replayable evidence.
This project demonstrates an autonomous drone navigation system using Jev. It showcases the capability to pilot a drone between two points in a simulated urban environment while actively avoiding obstacles during the flight path.
This project features a competitive game of Pong where Jev faces off against various LLM opponents. It demonstrates real-time interaction capabilities and game logic integration within the Jev ecosystem through a simple, interactive interface.
JEVLAB ARTTyped, confidence-aware agent skill routing with TypeSafe Jev.
JEVLAB ARTA 2-minute song composed by Jev-thoven using first principles.
chowder is a unified API layer that aggregates, normalizes, and serves your data from any source. One endpoint. Every ingredient.
Explore sponsorThis project demonstrates an integration where Jev is utilized to play Pokémon Red. It showcases a turn-based interaction model where the AI processes inputs to navigate the game environment and execute specific commands during gameplay sessions.
JEVLAB ARTJev Sift is a developer tool designed for agent-based classification and selective reading. It utilizes an agent plugin and MCP tool to help users filter and organize information streams efficiently.
JEVLAB ARTAn agent using headless Chromium to quickly navigate the web, demonstrated by browsing Wikipedia links efficiently.
This Python CLI tool demonstrates an experimental approach to semantic line searching using TypeSafe Jev and OpenRouter. It provides a lightweight interface for querying data without requiring complex runtime dependencies.
This demonstration showcases Jev processing 700 leads to predict outreach performance. The creator claims the system completes this analysis in 40 seconds, illustrating a potential workflow for rapid lead qualification and messaging optimization.

The AI-augmented control plane for cybersecurity — unify your security stack, quantify cyber risk in dollars, and run governed AI agents.
Explore sponsorThis repository serves as a sandbox for testing Jev, the System One model from TypeSafe. It provides a practical environment for developers to explore the model's capabilities through various experimental implementations and code samples.
This project showcases a Rubik's Cube prototype that utilizes Jev AI to perform self-solving maneuvers. The creator demonstrates an approach modeled after human problem-solving techniques instead of relying on traditional optimal mathematical algorithms.
This directory provides a curated list of systems utilizing Jev for Noul, Score, and Choice question formats. It serves as a central reference for exploring various implementations and applications built within the Jev ecosystem.
This repository provides a Go SDK for interacting with the TypeSafe AI API. It enables developers to integrate TypeSafe AI functionality directly into Go applications by providing structured client methods for API communication.
A developer project for the agents shelf, using MCP tool for Codex instead of listing all options.
Lemonpod gives founders one morning brief that combines GitHub, calendar, tasks, inbox, and other work signals into a text summary and audio update. Start the day knowing what needs attention without checking multiple apps.
Explore sponsorThis demonstration showcases how Jev can track and respond to user sentiment over time. It utilizes structured state management to maintain a persistent mood profile based on the tone of ongoing interactions.
Run a JSON Schema through TypeSafe's Jev API, and get JSON back.
This demonstration explores using inference engineering and scoring endpoints to guide open source models toward Jev-style decision-making. The creator showcases a workflow for implementing these behavioral adjustments within the model architecture.
TrustGate provides an independent media trust and safety gateway utilizing TypeSafe Jev. This tool demonstrates a framework for managing content moderation and safety protocols within decentralized media environments through structured, type-safe validation processes.
A Chrome extension using Jev to filter out AI-generated content from YouTube, caching results for efficiency.

Whale sharks drifting past. A quiet café after hours. Sixteen living worlds, just beyond your screen.
Explore sponsorA demonstration of Jev competing against OpenJev in a first-person shooter game.
RoboJEV is a developer project demonstrating a two-stage Jev control system for a robotic arm within the MuJoCo simulation environment. It serves as an experimental implementation for testing hardware-focused control logic.
JevArena demonstrates a competitive environment where two Jev agents engage in browser-based click games. This project explores the integration of Browser Use and TypeSafe Jev frameworks to facilitate autonomous agent interaction within web interfaces.
This local proxy utilizes TypeSafe Jev to dynamically route subagent requests to specific Claude models and effort levels. It is designed to manage specialized tasks independently while preserving the state of your primary chat session.
This project integrates Jev into the classic Chrome Dino game to provide real-time decision scoring. It demonstrates how external logic can influence gameplay mechanics by displaying live evaluation metrics during the runner sequence.
TypeSafe Jev plays original Civilization II in a browser, with live action probabilities. Experimental full-game harness.
Local A-share technical analysis using Jev Choice and Noul questions to assess direction and volatility across two horizons, with evidence for Buy/Hold/Sell assessments and no order execution.
TypeSafe's Jev plays Atari Pong. One typed Choice question per frame, no coordinates sent to the model.
This repository provides a comprehensive tracker for the Jev ecosystem, documenting over 220 cases. It offers confidence-graded entries that are automatically rescanned every three hours and includes a guide for API access.
This resource documents a matched-precision comparison between a private fine-tuned classifier and standard models. It provides insights into decision-making accuracy when evaluating specific classification outputs against custom-trained benchmarks.

Whale sharks drifting past. A quiet café after hours. Sixteen living worlds, just beyond your screen.
Explore sponsorThis repository demonstrates a Jev-style calibrated decision model built on the Qwen3.5-0.8B architecture. It explores the implementation of Choice, Score, and Noul decision frameworks within a compact language model environment.
This project provides a native macOS menu bar application for Notion. Built with Swift and SwiftUI, the tool aims to offer a lightweight interface for accessing Notion workspaces directly from the system menu bar.
This project demonstrates an AI-driven civilization simulation where TypeSafe Jev manages decision-making through typed, probabilistic, and auditable processes. It integrates with various LLMs via OpenAI-compatible APIs to handle high-level planning within the game environment.
TypeSafe Jev demonstration for new analyzation — experimenting with Jev for fast analysis of news and tickers.
This unofficial PHP SDK provides an interface for the TypeSafe AI System One API. The creator claims it maintains functional parity with official JavaScript and Python SDKs, though it is not affiliated with TypeSafe AI.

Guideless is an AI-powered platform that turns software workflows into video training guides.
Explore sponsorThis repository provides a reproducible benchmark comparing Jev and Claude Haiku 4.5 across 2,000 phishing emails. It evaluates performance metrics including accuracy, calibration, latency, and cost to assess model effectiveness in detecting malicious communications.
Jev automates the filling of prompt boxes by selecting appropriate agents, models, or computers for specific tasks.
Jev converts websites into native mobile apps by deciding the optimal build process and handling app store submissions.
This TypeSafe AI reference guide explains the re-ranking process for improving information retrieval accuracy. It demonstrates how to refine search results by re-evaluating candidate documents to ensure the most relevant content is prioritized for the end user.
Live verdict page: feeds Jev the day’s Florida Man, odd-news, politics and world headlines and asks all three primitives whether AI should kill us all, refreshed every ten minutes.

Menta es el software de gestión clínica impulsado por IA. Una plataforma todo en uno para la gestión administrativa y clínica de profesionales y clínicas
Explore sponsorA teleprompter of talking points that checks each one off as you cover it, using TypeSafe's Jev Score questions.
I tortured Jev into being a RISC-V CPU.
A community post describes using Jev to analyze news for stock investment opportunities.
This official resource provides a collection of reference demonstrations for TypeSafe AI. It serves as a central index for users to explore practical implementations and verified examples of the platform's core capabilities and architectural patterns.
Beat Jev is a game project featuring a penalty shootout mechanic. It demonstrates the integration of Render Workflows with a Postgres database to manage game state and user interactions within a web-based environment.
JEVLAB ART
Wallpets brings cinematic animal companions to your desktop with lightweight mouse-following behavior, idle play, and a growing pet catalog.
Explore sponsorAn open-source model for form-filling tasks with low latency.
A tool that improves end-to-end testing efficiency, completing tasks faster and cheaper than other models.
This project demonstrates an automated agent designed to identify and filter low-quality content as users scroll through their feeds. It aims to improve digital consumption by providing real-time classification of incoming information streams.
This project demonstrates a cost-aware LLM router designed to select the most economical model for a given query. It utilizes Jev for rapid classification, aiming to reduce operational expenses by avoiding unnecessary LLM calls.
This demonstration showcases Jev navigating a webpage by selecting specific candidate links to reach a defined goal. It illustrates the agent's decision-making process for link interaction within a browser-based environment.

The AI-augmented control plane for cybersecurity — unify your security stack, quantify cyber risk in dollars, and run governed AI agents.
Explore sponsorThis Chrome extension integrates Jev to analyze X posts for tone and sentiment. It allows users to perform a vibe-check on their draft content before publishing to ensure the message aligns with their intended communication style.
Jev trades 15-minute and 1-hour BTC, ETH, and SOL markets on Kalshi.
Explains Jev's advantages over older LLMs in decision-making and building.
A GitHub project that uses Jev to triage and route posts from the live Bluesky firehose with human oversight.
JEVLAB ARTThis project introduces a system for deploying agents directly onto digital canvases. The creator claims this approach bypasses traditional, slower LLM processing workflows to enable more rapid agent interaction and deployment within the interface.
This work-in-progress tool automates pull request reviews by applying plain-English Jev rules. It aims to streamline code quality checks by providing automated feedback based on defined project standards.
This tool utilizes Jev evaluation capabilities to rank Pi agent skills for specific tasks. It demonstrates a method for optimizing agent performance by matching specialized capabilities to user requirements through systematic assessment.
This Claude Code plugin integrates Jev to score review findings, debugging hypotheses, and design options. It aims to provide calibrated probability assessments rather than subjective opinions during the development process.
A simple Spring Boot 4 starter for TypeSafe Jev using Spring MVC and RestClient.
Jev Auto Router demonstrates a per-call routing system for Codex that dynamically selects GPT models and effort levels. It utilizes a local proxy to maintain tool loops while employing independent verification to confirm task completion.
Lemonpod gives founders one morning brief that combines GitHub, calendar, tasks, inbox, and other work signals into a text summary and audio update. Start the day knowing what needs attention without checking multiple apps.
Explore sponsorMagic Jev Ball is a tool designed to streamline GitHub pull request workflows. It utilizes Jev to evaluate CI status, diff size, and existing reviews, aiming to provide automated approval decisions within a reported timeframe of 200 milliseconds.
This official documentation provides the reference for the TypeSafe AI JavaScript SDK. It outlines the necessary methods and integration patterns for developers to implement TypeSafe functionality within their JavaScript-based applications.
This project provides a collection of interactive graphical modules including routing, a Tetris implementation, swarm simulations, and a gauntlet challenge. It serves as a practical demonstration of Jev-based application development and game logic implementation.
This Chrome extension utilizes TypeSafe Jev to identify and block native advertisements, sponsored content, and video ads. The project demonstrates a privacy-focused approach to content filtering by applying semantic analysis to browser feeds.
This community tutorial explores Jev capabilities, featuring practical demonstrations of browser automation and AI memory management. It provides a foundational look at how these agents interact with web environments and retain contextual information for tasks.
chowder is a unified API layer that aggregates, normalizes, and serves your data from any source. One endpoint. Every ingredient.
Explore sponsor