Top Picks

A tiny desktop utility that shows remaining GitHub Copilot credits in the macOS menu bar or Windows system tray, using GitHub CLI login and no token storage.

Open-source persistent workspace for AI agents to work with market data, run rules, compare results with web sources, keep journals, and optionally place paper orders.

Hapsland is a developer tool that gives coding agents realtime semantic review on code changes, with privacy-aware request scoping and feedback loops back to the agent.

A prototype that turns a Nokia 110 4G into a chat-driven AI device, with actions like calling, alarms, torch control, battery checks, and SIM data queries.

HDTP is a protocol specification for agent-to-agent communication on behalf of people, centered on certificates, contact approval, per-caller permissions, and sealed end-to-end calls.

Portia is a copilot for data work that connects to warehouses or local files, profiles tables, answers questions, draws charts, and builds dbt-shaped SQL with step-by-step review. It also ships as a Claude Code plugin and keeps a journal of decisions, checks, and lineage.

An MCP server and API that checks documents for hidden or invisible text before an AI assistant reads them.

OpenAI API documentation for the Decisions API, which checks conditions, selects from fixed options, and scores text or images against a rubric.

Genjutsu AI is a video-to-video tool that keeps the original subject and motion while restyling the scene, objects, or look around it. It supports motion transfer, object swaps, multiple resolutions, and one-time credit pricing.

A reproducible GitHub research repo comparing Codex reasoning-token usage between API and ChatGPT Pro access across 960 completed responses, with code, data, and verification steps included.

A small open model trained to turn English requests into Bash commands, with published training data, GGUF weights, and benchmark results.

Browser-based RAG app that runs entirely client-side, using BM25/Lunr.js for retrieval and direct LLM provider APIs for answering questions without a backend.

A deterministic security scanner and adversarial flight simulator for AI agents, built for red-teaming LLM workflows and MCP-compatible setups.

Graphene is a data analytics framework built for coding agents, with a CLI, SQL-based semantic models, and Markdown pages for dashboards and reporting.

Open-source LLM routing and cascade engine that sends prompts from local Ollama models to cloud and frontier models based on task complexity, cache hits, and quality gates.

DoorDash’s corporate ordering connector lets companies connect AI tools or in-house agents to search DoorDash, build carts, place orders, and track delivery for team lunches and office supplies.

An isometric, Theme Hospital-inspired control panel for an AI team, with drag-and-drop task assignment, a journal-backed backend API, and dispatcher scripts for running models.

A long-form research article on why browser-based AI agents fail on real websites and how accessibility engineering and task-level tests can reveal the breakpoints.

A CUDA C++ project for training a tiny byte-level GPT quickly, with build, run, checkpoint, logging, and test instructions in the repo README.

Viamour is a travel and community app for finding friends, joining local meetups, and booking curated experiences in new cities.

Research code for improving agent harnesses through regularized recursive self-improvement, with evaluation, selection, pruning, and domain-specific adapters.

A GitHub repository for a 7-node ESP32-S3 cluster that runs a small LLM using 1.58-bit BitNet-style quantization over an SPI daisy-chain.

A set of Claude Code skills for analyzing chess games with Stockfish-backed annotations, HTML viewing, transcription from think-aloud audio, and narrated video output.

A local-first AI calculator that explains arithmetic step by step in a browser-based workspace. It supports percentages, powers, roots, parentheses, and follow-up math while keeping conversations stored on the device.