Recently Evaluated

arXiv paper on distilling multi-agent debate into a single LLM with a two-stage fine-tuning pipeline. The abstract reports lower token use, comparable or better benchmark performance, and an analysis of agent-specific activation subspaces, with code linked from the page.

Cloud cost awareness for AI coding agents and IDEs, with real-time pricing for IaC changes, cost tradeoff suggestions, tagging fixes, and policy-aware guardrails.

Lowfat is a lightweight CLI tool that filters noisy command output before it reaches an AI agent, with plugins, shell integration, and usage stats for token savings.

An open-source reference implementation for autonomous vulnerability discovery and remediation with Claude. It includes Claude Code skills for threat modeling, scanning, triage, patching, plus a harness for running a recon → find → verify → report → patch pipeline.

Piece is a mobile app that uses AI to help couples settle small everyday arguments by hearing both sides, weighing the claims, and delivering a verdict in a chosen tone.

A blog post describing a small reinforcement-learning agent trained with PPO to play and beat a Pokelike/Pokerogue-style game, including the input representation, model architecture, and training approach.

CLI for installing, switching, and managing collections of agent skills and agent-specific configurations across multiple coding tools.

h5i is an AI-aware Git sidecar that records agent activity, reasoning traces, prompts, model metadata, and review signals in Git refs so teams can hand off work between coding agents and surface PR review briefs.

A free AI-readiness audit tool for Shopify catalogs that scores product visibility for AI shopping agents and flags catalog issues like weak titles, short descriptions, missing attributes, and sparse images.

Visr is a CLI tool that records full agent terminal trajectories, including commands, output, retries, and fixes, then returns a transcript link for reuse across sessions.

Hydron is a hardware-native AI coding assistant for embedded engineers. It offers spec-cited code generation, datasheet-based answers, and debugging workflows tied to real hardware, with support for VS Code and CLI use.

Swift package for building provider-agnostic AI agents on Apple platforms, with tools for runtime permissions, hooks, context compaction, session state, validation loops, and traceable execution.

A black-box benchmark report on how AI-generated tests detect functional bugs in live APIs across 20 scenarios and 7 systems.

Local-first AI memory layer for LLM apps with persistent knowledge graph storage, entity extraction, semantic retrieval, and API endpoints for ingesting and retrieving context.

Google’s Gemma 4 12B is a multimodal model for local and developer-focused AI use, with native audio and vision support, agentic workflows, and broad tooling support.

AI-native workspace for writers, bloggers, and lifelong learners. It combines source collection, an idea map, AI-assisted exploration, and a focused writing editor designed to support long-form thinking.

Open-source Go runtime for autonomous multi-agent systems. Swarm models work as durable state machines, persists state, supports replay/fork, and integrates with multiple LLM backends.

A Claude Code skill for multi-agent code review that produces findings.md with severity, impact category, verification artifacts, and fix directions.

Remote control for Claude Code and Codex on your own machine, with browser and Telegram access, persistent sessions, and BYO AI accounts/keys.

An API for mapping hex values, palettes, and images to named archive colours, with deterministic colour metrics, accessibility checks, cultural risk notes, and MCP support for agents and chat tools.

A demo for running the Odysseus self-hosted AI workspace on throwaway islo.dev microVMs in one command, with both a clickable live app path and an agent tool for running code on disposable boxes.

Browser extension that scans pastes for credentials and PII before they reach LLM chat sites, with local on-device inference and no network calls.

kLex is the reference runtime for FROG, a typed language built for native tooling, browser apps, pipelines, and concurrent utilities. The repo shows install steps, language examples, a browser playground, GUI support, and a security-scanner demo.

AI-powered knowledge base for Terraform modules that supports semantic search, agentic composition, audit checks, and optimization from indexed repositories and conventions.