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
Nvidia CEO on AI Economic Impact
Jensen Huang describes AI as a transformative force capable of reshaping the global economy. He argues that AI factories will drive industrial growth and create significant demand for physical labor.
threads.comrust-sysone
This unofficial Rust client provides a type-safe interface for interacting with System One TypeSafe AI. It demonstrates a fluid API for constructing requests and handling typed responses, including token usage and cost estimation.
github.comdemo-symfony-typesafe
This project demonstrates how to use Jev and Symfony AI to automatically categorize public support requests. It showcases asynchronous processing, rate limiting, and real-time dashboard updates using Symfony components to classify intent, urgency, and bug reports.
github.compi-typesafe
pi-typesafe integrates the Jev judgment model into the Pi agent environment. It provides a tool for structured data evaluation, allowing agents to perform classification, scoring, and boolean checks with calibrated probabilities instead of prose.
github.comJev DSPy Lab
Jev DSPy Lab provides tools for recording, replaying, and benchmarking TypeSafe AI decisions within DSPy pipelines. It enables developers to measure calibration, selective risk, latency, and modeled costs using deterministic offline analysis.
github.comjev-usecases
This repository provides Python implementations for TypeSafe AI's Jev model, demonstrating how to use calibrated probabilities for automated, typed decision-making in software. It includes various use cases like security triage, invoice processing, and guardrails for generative language models.
github.comjevymarket
jevymarket is an experimental trading bot for Polymarket that uses the Jev decision model to calculate probabilities. It combines web-based research with Jev's structured analysis to identify and execute trades based on configurable edges.
github.comlkclean
lkclean is a Chrome extension that filters LinkedIn feeds by using Jev to classify posts. It collapses engagement bait and irrelevant content while providing transparent explanations for every decision made by the model.
github.comjev-test-filter
jev-test-filter uses Jev to analyze git diffs and identify which tests are affected by code changes. It automatically generates and executes the necessary filter arguments for test runners like vitest, Playwright, cargo test, and go test.
github.comslop-filter
This Chrome extension identifies and hides AI-generated content on social media platforms. It uses TypeSafe Jev to score posts based on specific linguistic patterns, allowing users to tune detection weights according to their own personal judgment.
github.comjev-reranker
jev-reranker is a Python library that uses the Jev API to filter and reorder search results for RAG pipelines. It provides configurable thresholds to remove irrelevant documents and supports both synchronous and asynchronous requests for efficient context management.
github.comjev-minesweeper
This project demonstrates using the Jev AI model to solve Minesweeper. It employs speculative fan-out and confidence-gated routing patterns to query the model for mine probabilities, comparing its performance against a deterministic solver.
github.comLLMtoJev
LLMtoJev is a tool that analyzes LLM prompts to identify bounded decisions suitable for migration to Jev primitives. It helps developers extract classification, scoring, and boolean tasks from generative prompts to improve efficiency and structure.
github.comtypesafe-go
This repository provides Go utilities for the TypeSafe API, including a natural language grep-like CLI, a Kubernetes log analysis plugin, and a standard-library-only SDK. It demonstrates how to integrate TypeSafe services into Go-based workflows for automated data querying and log inspection.
github.comHearth Rental Search
Hearth is an autonomous agent that uses TypeSafe Jev to navigate multiple rental marketplaces. It performs browser-based searches to aggregate listings into a single shortlist, focusing on reading and reporting data without executing transactions.
github.comjev-rs
jev-rs is a Rust-based engine that provides TypeSafe-compatible System One judgments from LLMs. It extracts log-probabilities for noul, choice, and score tasks without generating text, offering tools for local calibration and evaluation of model confidence and decision accuracy.
github.comjev-voice-control
This macOS menu-bar application translates spoken commands into structured actions using Jev, a System One model that provides typed decisions rather than generated text. It combines deterministic local execution with AI-guided UI interaction to control applications and system settings.
github.comJev AI System One
Jev AI is a System One model designed for software integration, providing typed, calibrated decisions like choices, scores, or probabilities. It aims to replace prose-based LLM outputs with structured data to enable direct, error-free programmatic decision-making.
github.comjev-grand-prix
This project demonstrates an F1 racing simulation where Jev acts as the driver by selecting racing lines and pedal inputs. It uses a hybrid approach where code handles physics execution while Jev learns optimal cornering speeds over multiple laps.
github.comJevmoji
Jevmoji is a tool that uses TypeSafe's Jev model to score and retrieve relevant emojis based on user input. It demonstrates how to integrate AI-driven semantic scoring with local keyword matching for emoji discovery.
github.comjev-practice-speed
This WebGL demo features a Speed card game where users play against a CPU powered by TypeSafe AI's Jev. It visualizes Jev's real-time decision speed and accuracy, using a mechanical validator to ensure game rules are followed.
github.comjev-information-extraction
This project demonstrates how to use the Jev model to extract information from PDF documents. It provides an interactive interface for ranking text chunks based on user-defined questions, visualizing evidence directly on the original document pages.
github.comjev-connector
This WordPress plugin integrates the TypeSafe System One API, enabling developers to query content for structured data like probabilities, categories, and scores. It leverages the WordPress Connectors API for secure key management without using generative text features.
github.comtypesafe-ai-go
This community-maintained Go SDK provides a structured interface for the TypeSafe AI System One evaluation API. It features fluent builders, typed question definitions, and built-in retry logic with exponential backoff to facilitate reliable interaction with the evaluation service.
github.comHermes and Jev play Minecraft
This project demonstrates an agent architecture where Hermes plans and Jev selects bounded actions to play Minecraft. It reproduces a successful Ender Dragon speedrun, showing that harness-side validity checks and specialized controllers outperform general LLMs in efficiency and cost.
github.comJev Realtime Trading
This desktop application demonstrates a paper-trading system that uses TypeSafe's Jev model to analyze live market data. It executes simulated trades based on predefined rules, providing a visual interface for monitoring agent decisions and performance without using real capital.
github.comLaya vs Jev AI Arena
This tool provides a comparative testing environment where Laya and Jev models compete in Snake and fighting games. It demonstrates how different AI models handle real-time decision-making tasks under identical conditions and performance constraints.
github.comjev-benchmark
This repository provides a reproducible benchmark for TypeSafe AI's Jev, evaluating its performance on agent tool-call risk classification. It specifically tests accuracy, latency, and the reliability of confidence scores across clear, ambiguous, and adversarial task scenarios.
github.comwatfile
watfile is a command-line tool that automates file organization by classifying documents into folders using AI. It supports cloud-based classification via TypeSafe AI Jev or local processing with Laya models to sort files by content.
github.comtypesafe-jev-mcp
This MCP server integrates the TypeSafe Jev model into AI agents as a typed evaluation tool. It demonstrates how to process state and structured questions to return categorized outputs like choices, scores, and probabilities for automated decision-making workflows.
github.comawesome-jev
This directory serves as a community-curated hub for the TypeSafe Jev ecosystem. It aggregates official SDKs, agent-based demos, and experimental projects that utilize System One models for structured decision-making tasks across various applications.
github.comTypeSafe API Response Documentation
This documentation details the SystemOneResponse structure within the TypeSafe Python SDK. It explains how the API returns structured answers, confidence scores, and token usage metadata for various question types.
docs.typesafe.aiTypeSafe AI Questions API
This documentation outlines the Python SDK for TypeSafe AI, focusing on defining yes/no, choice, and score-based questions. It demonstrates how to structure state and criteria using objects or dictionaries to facilitate structured interaction with AI models.
docs.typesafe.aiTypeSafe AI SDK RetryPolicy
This documentation details the RetryPolicy class for the TypeSafe AI Python SDK. It explains how to configure retry attempts, backoff strategies, jitter, and specific HTTP status codes to handle transient network errors and API connection issues effectively.
docs.typesafe.aiTypeSafeClient Sync API
The TypeSafeClient provides a synchronous interface for interacting with the TypeSafe AI API. It allows developers to configure authentication, retry policies, and timeouts while facilitating model listing and structured data queries.
docs.typesafe.aiTypeSafe SDK Exception Handling
This documentation outlines the exception hierarchy for the TypeSafe Python SDK. It details how developers can programmatically handle various API errors, including HTTP status codes, connection failures, timeouts, and response validation issues during integration.
docs.typesafe.ai
