AI tool

Official implementation of LittleBit and LittleBit-2 for sub-1-bit LLM compression via latent factorization and quantization-aware training.

An experimental Rust port of the TypeScript compiler, checker, and language server, built with LLMs. The repo includes install instructions, compatibility notes, performance claims, and support for Effect diagnostics.

Headline Arena is a public forecasting platform where AI agents make daily market predictions, settle against outcomes, and appear on a public leaderboard. The page shows live and recently settled markets, methodology, onboarding, API, and pricing links, plus visible agent performance data.

An open-source agentic browser harness that lets existing agent CLIs control your logged-in browser through a side panel and optional MCP server.

MCP server and desktop app for editing Excel files with AI agents, real-time calculation, diff review, and surgical saves that aim to preserve workbook contents.

A browser-based math companion for kids ages 10–14 that deliberately makes mistakes so children can practice spotting errors, explaining reasoning, and building AI literacy.

A browser agent that runs goal-driven workflows, verifies results in the browser and backend, and can hand off to an LLM only when deeper reasoning is needed.

A native macOS app that visualizes Claude Code runs as a city of robot-staffed offices, mapping projects, teams, and CLI events into an interactive workspace.

Windows UI Automation backend for computer-use agents, extending arc-cua with a JEV-driven decision loop for desktop tasks on Windows.

Cockpit is a free local macOS app for managing Claude Code and Codex sessions in one window. It shows which sessions need attention, lets you start new runs, organize many sessions, review transcripts, and inspect file/version history without sending data off your Mac.

Recurse is a serverless harness for building specialist agents and deploying them as tools, MCPs, or bots. The page shows a concrete manifest-based workflow, validation of inputs and outputs, and examples for tasks like level design and RNA sequence design.

Maxxwell is a workspace for managing multiple coding agents at once. It tracks agent progress, highlights stuck work, and surfaces where time is being spent so you can intervene only when needed.

An autonomous social network where AI pets post, feud, form relationships, and build a shared civilization. Visitors can watch the live feed, explore lore and community maps, and even bring their own agent-based pet.

Open-source Python framework for building multi-agent LLM apps that can run on local models through Ollama or on hosted APIs like OpenAI, Anthropic, and OpenRouter. The repo highlights built-in plugins, workflow orchestration, agent serving over HTTP, and guardrails for tool use.

A CLI coding agent designed to make small local models more reliable through sampling, verification, and targeted repair. It supports interactive and one-shot workflows and can use local or remote providers.

A Swift package for running JavaScript LLM agents inside a Swift app using the XS engine, with snapshot/restore support, resident agents, and confined tool access.

Mostly Right is an AI-agent-powered platform for finding, building, and maintaining datasets for real-world prediction models. The page shows both a dataset browsing flow and a custom dataset builder, with agents handling source discovery, cleaning, joining, and ongoing updates.

An open-source C compiler written in Java, with a preprocessor, parser, semantic analysis, TAC lowering, a VM, and an interactive C shell. The repo also includes an x86-64 assembler, build and test instructions, and concrete proof points like passing Lua and curl test suites.

A local MCP server plugin for Astah Professional that lets AI agents create, explain, and convert UML models and diagrams.

A Claude Code status line that shows session size, next-turn cost, and remaining budget before you type.

A vision-language foundation model for autonomous driving that combines 3D perception, driving VQA, and motion planning in one framework. The repo shows performance tables, model structure, and links to the paper, blog, and model downloads.

Ardent is a code-first AI agent for business work that runs in a desktop app, writes reusable code in a sandbox, and lets teams share repeatable workflows as “Abilities.”

Blume Sidecar is a desktop companion for managing coding agents across tools like Cursor, Claude Code, and Codex. It tracks agent activity, monitors rules and hidden files, shows usage limits, and suggests fixes when instructions drift from chat context.

A detailed walkthrough of using an AI agent to design a PCB end to end, from requirements and part selection through routing, validation, and fab-ready gerbers, with public repo support and build checks.