ZeroFS - ZeroFS serves S3-compatible buckets as POSIX filesystems over NFS and 9P, or as raw block devices over NBD.

Fast, collaborative live terminal sharing over the web

Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.

The "Missing GitHub Status Page" -- a Flat Data attempt at historically documenting GitHub statuses

Measuring frontier coding agents on original, long-horizon engineering tasks

TokenSpeed is a speed-of-light LLM inference engine.

How much experts do we need to serve a model?

Fast, lossless LLM inference via dual-view diffusion decoding.

DFlash & TurboQuant in llama.cpp with up to 3x faster generation and 7.5x more KV cache in same VRAM

Monoscope lets you ingest and explore your logs, traces and metrics. We store these in S3 compatible buckets. Query in natural language via LLMs.

llama.cpp fork with TurboQuant WHT-rotated KV cache & weight compression + Gemma 4 MTP and Qwen 3.6 NextN speculative decoding (+30-50% throughput).

Hierarchal Agent Loop Optimizer

A filesystem designed for agents, with SOTA retrieval, automatic memory profiles, sync engine. Drop any file type (pdf, images, videos), and grep through them.

🤗 ml-intern: an open-source ML engineer that reads papers, trains models, and ships ML models

🐚bash/POSIX-compatible shell implemented in Rust 🦀

Specification and Tools for Makefile-formatted Agent Skills.

Reverse proxy for monitoring and debugging local LLM agents (Ollama). Real-time dashboard, request logging, and performance metrics in a single binary

Warp is an agentic development environment, born out of the terminal.

Dataset of hackable TerminalBench-style tasks and exploit trajectories

Structured Chain-of-Thought

VS Code extension that follows file changes in real-time, automatically opening editors and scrolling to edits. Perfect for watching CLI-based coding agents work.

FlashKDA: high-performance Kimi Delta Attention kernels

👩‍🚒 Good content deserves good paper.

A theoretical reconstruction of the Claude Mythos architecture, built from first principles using the available research literature.

Domain-fronted HTTP/SOCKS5 proxy tunneling traffic through Google Apps Script with MITM TLS interception, HTTP/1-2 multiplexing, and DPI evasion.

Instant, Concurrent, Secure & Lightweight Sandbox for AI Agents.

Library for reducing tail latency in RAM reads

Give Claude Code a memory that evolves with your codebase. Hooks automatically capture sessions, the Claude Agent SDK extracts key decisions and lessons, and an LLM compiler organizes everything in…