Recently Evaluated

Focal Prompt is an analysis lab for studying how prompts and prompt sections influence model behavior. It includes focus labeling, prospective and retrospective focus prediction, baseline output sampling, ablation tests, batch analysis, and an agent builder.

An in-memory Tree-sitter AST pruning and context firewall MCP server for coding agents like Cursor and Claude. The repo includes installation steps, diagnostics, benchmarks, and a clear usage path for local setup.

Weekly tracking for whether ChatGPT, Perplexity, Google AI Overviews, Gemini, and Claude cite your site, name your brand, or leave you out. The page shows pricing, the report structure, example outputs, and how the system works.

A blog post introducing Orca, an open-source agent development environment for running multiple coding agents in parallel with git worktrees, diff-based review, browser previews, and mobile notifications.

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.

A LangGraph-based workflow that generates, benchmarks, validates, and iterates CUDA kernels using a C++/CUDA harness. It targets workload-specific optimization with support for NVIDIA research lookup and optional Nsight Compute profiling.

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.

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.

A web tool that turns a public website into an MCP server so AI agents can search and read its content via a generated link.

Free curriculum for learning AI agents, LLM evaluation, and benchmarking, with 42 lessons, interactive labs, runnable JavaScript, and saved progress.

CraftCX is a support QA and intelligence tool for AI support teams. It reviews AI-handled conversations, scores support quality, surfaces customer issues, and connects findings to Intercom, Front, HubSpot, and Gorgias.

A local gateway for Claude Code that logs tool calls, enforces approval rules, and trims noisy tool output to save tokens.

A Claude Code plugin and CLI that tracks usage-limit reset times, writes handoff notes before limits hit, and resumes headless or subagent work after the reset.

A research project and proxy for Claude Code that chooses per-step reasoning effort with Jev while preserving the prompt cache. The repo includes benchmark results, a real-session measurement, and notes on how the mechanism works.

A Mac notch app that surfaces Claude Code and Codex approval prompts in the menu bar/notch so developers can approve or deny requests without switching back to the terminal.

Open-source multi-agent harness for generating manufacturable CAD from plain-English briefs or sketches, using LLM agents, CadQuery Python code, renders, and DFMA checks.

A GitHub repo benchmarking TypeSafe’s Jev against Sonnet, GPT-5 nano, and local models on 770 Reddit AITA verdicts, with Brier scores, latency, and cost comparisons.

arXiv paper on constraining LLM decoding with incremental parsing to reduce invalid SQL and improve text-to-SQL results.

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.

A Windows tray app for monitoring Claude Code usage limits, remaining context, and history, with alerts, charts, and local storage.

An MIT-licensed terminal coding agent for the Step ecosystem. It reads code, edits files, runs tests, supports MCP servers and plugins, and includes built-in static site publishing.

A text-first, hands-on guide to agentic AI for practicing engineers, with runnable local-first code samples and a module-based structure covering LLM basics, tool calling, RAG, memory, multi-agent patterns, evaluation, safety, and enterprise topics.

ToolHub is a self-hosted open-source platform that turns scripts into AI tools and organizes them in a hierarchical tree for agents to browse and execute. The page shows architecture notes, execution flow, screenshots, and a live demo/web client reference.