Submit a find
한국어
日本語
Español
Français
Deutsch
简体中文
繁體中文
Português
Bahasa Indonesia
ไทย
Italiano
Tiếng Việt
Türkçe
Polski
Русский
Back to the labJEVLAB / FIELD NOTES
Fresh from the lab
A dated reading list of newly added resources and JEVLAB NEWS. Come back for what changed, or follow the RSS feed in your own reader.
Remember this visit on this browser to highlight new additions next time. No sign-in or cross-device tracking.
Dates show when an item became available in JEVLAB, in UTC, not when its original source was published. Translation-only edits do not count as new resources.
36
casse-brique-typesafe
This project demonstrates a Next.js brick breaker game where the paddle movement is controlled in real time by the TypeSafe AI Jev model. The creator reports that the application was developed using Claude Code.
github.comJev plays Clash Royale
This demonstration shows Jev interacting with the mobile strategy game Clash Royale. The footage captures the agent navigating game mechanics to complete and win a match, illustrating its capability to process real-time visual inputs and execute tactical decisions within a gaming environment.
x.comjev-model-router
This tool demonstrates a cost-optimized model router for OpenRouter using Jev. It replaces static model lists with a dynamic, live catalog scorer to select appropriate models for specific tasks.
github.comClaude Code Context Compaction
This demonstration explores integrating Jev with Claude Code to manage context compaction within agent workflows. It highlights a practical approach for optimizing data handling during automated coding tasks, though the effectiveness of these results remains subject to user verification.
x.comjev-cookbook
This collection provides practical, tested recipes for implementing the Jev decision model via OpenRouter. It demonstrates workflows for tasks including PII detection, database indexing, file organization, and browser agent automation.
github.comjev-spec
jev-spec is a utility designed to verify generated code against predefined requirements. This tool aims to assist developers in ensuring that AI-produced outputs adhere to specific structural or functional specifications during the development process.
github.comjev-libero
This project demonstrates fine-grained robot control using Jev. It provides physics-based previews and configurable tasks within the LIBERO framework to help users experiment with robotic manipulation and simulation environments.
github.comterrarium
terrarium is a sandbox environment where a TypeSafe System One model operates a creature's controls. The project demonstrates how code-driven logic can be used to simulate and interact with a virtual world.
github.comJob Risk Analyzer
Job Risk Analyzer provides a CLI and REST API interface to evaluate professional roles. It uses Jev to score occupations based on their potential exposure to AI-driven workforce displacement and their relative resilience in an evolving labor market.
github.comJev-Moderation-Bot
This Discord bot utilizes Jev to analyze messages and metadata in real-time. It demonstrates a progressive escalation system designed to identify and mitigate phishing, spam, and social engineering attempts within server environments.
github.comtrade-jev
This project demonstrates the application of Jev for automated trading strategies. It provides a framework to backtest buy, sell, and hold decisions using NQ L10 order-book data to evaluate model performance in simulated market environments.
github.comjevlike
This project demonstrates a small model trained to select from dynamic text options using single-pass probability assignment. It includes experimental implementations for Doom, chess, and Wikispeedia to showcase the model's decision-making capabilities in varied environments.
github.comruby_decision_model
This repository provides a Ruby client library designed to interface with decision models, specifically targeting Typesafe Jev. It serves as a practical tool for developers looking to integrate structured decision-making logic directly into their Ruby applications.
github.comcompact-adviser
This agent plugin utilizes Jev to identify optimal moments for session compaction. It aims to assist developers in managing context more efficiently by providing automated guidance on when to consolidate session data.
x.comtypesafe-ui
This repository provides a collection of Shadcn-style reusable components and interface blocks designed for TypeSafe AI integration. It demonstrates how to implement consistent UI patterns for AI-driven applications using established design standards.
github.comJev Prompt Box Router
This project demonstrates an automated interface that populates prompt boxes by dynamically selecting the most suitable agents or models for a given task. It showcases a workflow for streamlining user input through intelligent routing.
x.comshady-town
Shady Town is a social-deduction party game designed for living room play. This project demonstrates how a TypeSafe Jev can act as an automated moderator to facilitate gameplay and manage interactions between participants.
github.comjev-cvss
This project demonstrates the use of TypeSafe Jev to automate CVSS scoring. It provides a structured approach to parsing vulnerability descriptions and generating standardized severity scores based on the provided input data.
github.comtypesafe-vs-deepseek
This project provides a side-by-side comparison of TypeSafe and DeepSeek-flash. It evaluates performance metrics including speed, token usage, cost, and accuracy across tasks like invoice extraction, email classification, and reranking.
github.comtypesafe-sdk
This repository provides a Ruby client library designed for interacting with the typesafe.ai platform. It serves as a programmatic interface for developers to integrate Jev-related services directly into their Ruby-based applications.
github.comjev-recall
This project demonstrates a method for filtering AI assistant memories using Jev. It focuses on retrieving information based on relevance rather than simple resemblance to improve the accuracy of context-aware interactions.
github.comJevForm
JevForm is a dynamic form tool that utilizes Jev for conditional logic. It demonstrates how to integrate adaptive UI elements using Vercel JSON-render to create responsive and interactive data collection interfaces.
x.comJev as a Reranker Evaluation
This experiment evaluates Jev as a search reranker, finding limited performance gains over standard vector retrieval. The creator reports mixed results that vary significantly based on the specific evaluation methodology applied during testing.
x.comJev Session Replay Debugging
This tool utilizes Jev to analyze user session replays for identifying software defects. It demonstrates an automated workflow designed to generate bug fix pull requests, aiming to streamline the debugging process for development teams.
x.comJev Decision-Making Framework
This resource introduces Jev as an AI-native system designed for decision-making. It demonstrates how the framework moves beyond simple text generation by defining potential outcomes and programmatically selecting the most appropriate path for autonomous operations.
x.comCrowdcheck
Crowdcheck is a tool designed to evaluate social media posts by simulating feedback from 10,000 synthetic personas. It aims to help users gauge potential audience reactions and refine content before publication.
crowdcheck-ai.vercel.appwindows-save-token-jev-setup
This resource provides a setup guide for configuring save-token-jev on Windows. It demonstrates how to implement PreCompact and SessionStart hooks to manage token security, native compression, and the isolation of legacy content within Jev environments.
github.comtypesafe
This library provides an idiomatic Elixir client for interacting with the TypeSafe AI API. It enables developers to integrate TypeSafe AI services into their Elixir applications using native language patterns and structures.
github.comconstruct-auto-classifier
This tool provides an effect-based safety gate for AI coding agents by filtering shell commands. It uses structural rules combined with Jev or chat models to evaluate command safety, aiming to prevent the execution of dangerous operations.
github.comAEPD Corpus Analysis
This experiment compares Jev against manual regex for processing 544 public data-protection resolutions. The creator reports a 98.2% agreement rate between the two methods, highlighting the potential for cost-effective automated document analysis.
x.comgpt-vs-jev
This tool provides a side-by-side comparison of language outputs generated by GPT models against structured Noul decisions. It demonstrates how Jev-based logic processes inputs differently than standard probabilistic language models for specific decision-making tasks.
github.comArbitrary classification as a primitive
This resource explores how Jev facilitates flexible, type-safe classification by treating it as a fundamental programming primitive. It demonstrates a conceptual approach to integrating classification logic directly into the development workflow for improved system reliability.
x.comJevmaxxing with Notte
This project demonstrates Jev, an agent designed to automate browser-based tasks. It functions by dynamically selecting and executing actions within a defined action space to navigate and interact with web interfaces.
x.comJev Gomoku Matchup
This experiment demonstrates a Gomoku match between two Jev instances. It provides source code and timing logs to illustrate how Jev agents interact and process game state during competitive play.
zenn.devPlayJev
PlayJev demonstrates a 0.8B parameter model capable of playing ten browser games. The system operates by calculating move probabilities in a single forward pass without generating text, showcasing a specialized approach to game-playing agents.
github.comyoshi
yoshi is a context-pruning proxy designed for Claude Code and Codex. It utilizes Jev to evaluate and retain only necessary history, aiming to optimize token usage for AI coding agents.
github.com
