Newly Accepted

Google Developers Blog post about integrating DFlash, a diffusion-style speculative decoding framework, into the vLLM TPU ecosystem to improve LLM serving speed on TPU v5p.

A GitHub repo for evaluating agentic AI pipeline systems, with guidance for defining metrics, building eval cases, running repeatable tests, and tracking regressions.

A metadata-first trust control plane for authorized security workflows, evidence retention, release trust, and business-flow proof. The repository includes role-based docs, quick-start commands, safety boundaries, and release-trust materials.

A ~1000-line C workflow runtime where an LLM can modify a DAG during execution using a restricted four-verb grammar, with every change written to an append-only audit log for replay and inspection.

Open-source agile planning and orchestration layer for AI coding agents, with bidirectional sync to Claude Code, AI-generated user stories, sprint planning, and a self-hosted CLI workflow.

An Apple Silicon–optimized inference build of Bonsai 1.7B with custom Metal kernels, benchmark results, quick-start instructions, and a bundled OpenAI-compatible server.

A hands-on workshop repository that walks developers through building a small GPT-style language model from scratch, including tokenization, transformer blocks, training, and text generation.

A macOS companion overlay for Claude Code that shows an animated pet/sprite based on Claude Code activity. It includes a CLI, hook integration, a daemon, and a web-based atlas validator/renderer for Codex-pet-compatible pets.

Sandboxed bash for agents with structured command output and full filesystem state capture after each run.

Lightweight AI inference engine written in Rust for embedding local LLMs in mobile and desktop apps, with bindings for Python, Flutter, and React Native.

An open-source experiment that adds a small zero-initialized overlay layer to a frozen GPT-2 so its behavior can be adjusted at inference time without retraining the base model.

Aurra is memory infrastructure for AI agents, with bi-temporal versioning, source-cited memory history, and an automatic supersession feature that uses an LLM classifier to decide when new facts replace old ones.

faz is a safety layer for AI agents that query databases. It sits between the agent and your data, adding RBAC checks, SQL/AST inspection, injection detection, guardrails, audit logging, and MCP/REST access.

An open-source project for making AI research agents legible, auditable, and traceable through behavioral code and event logs.

Open-source security scanner for AI agents, MCP servers, and agent skills. It auto-discovers installed agent components and checks them for prompt injection, tool poisoning, secrets, malware payloads, and related risks.

A general-purpose MCP bridge for Ableton Live that lets AI agents inspect and control Live sets, create and edit clips, add tracks and devices, analyze audio and harmony, and work with Max for Live audio taps.

A CLI proxy that lets Claude Code run against DeepSeek V4 Pro, OpenRouter, Fireworks AI, or Anthropic-compatible backends while keeping the same agent workflow.

An autonomous neuro-symbolic formal verification engine for Lean 4. The repository describes a proof workflow that generates tactics and then checks them deterministically in Lean, with example proofs, verification steps, and telemetry outputs.

Fabrica is a Rust-based terminal coding agent with a full TUI, multi-provider model support, and built-in file and shell tools for autonomous coding tasks. It also supports ACP server mode for programmatic control by compatible clients.

A Node.js proxy server that rotates API keys for LLM APIs like OpenAI and Gemini when rate limits are hit. It also includes an admin panel, request logging, provider management, and a Telegram bot for chatting with configured models.

AI-powered trade signal product for retail traders that turns financial news into ranked trade plans with entries, stops, targets, and tracked accuracy.

Semble is a local-first code search tool for agents. It offers natural-language search, an MCP server, bash/AGENTS.md integration, and a Python API, with claims of faster indexing and much lower token usage than grep+read.

A public leaderboard that ranks GitHub, GitLab, and Bitbucket repos by how friendly they are to AI coding agents such as Claude Code, Cursor, Devin, Codex, Gemini, Aider, and OpenHands.

Open-source AI sandbox infrastructure for running code, browsing the web, and isolating agent workflows inside disposable microVMs.