01 / MACHINE INTELLIGENCE
Models you can run. Agents you can inspect. Tools you can build on.
← All topicsServe large Qwen models fast on the GPUs you actually own. Qwen3.8-27B on a single 24 GB card with vLLM: 127 tok/s single-user (381 when the answer quotes the prompt), ~1,035 tok/s at 64 concurrent, 150k-262k context. vLLM patches, requant pipeline, benchmarks.
Run a hive mind of agents to focus on a specific goal
Route to the cheapest model in claude code for your task using jev-router
The game engine for AI agents. Give your AI superpowers for game dev + A skills framework that just works + easy install and setup ☀️
DevSecOps for the AI-era CI/CD pipeline. Catches the bugs your AI coding assistant introduces — before they reach production.
DELLA-Merging: Reducing Interference in Model Merging through Magnitude-Based Sampling
Monid - OpenRouter for agent tools.
Actually Useful Agent Observability.
Open the project to explore the idea and the code.
Pane tracker: all your AI plans in one Windows tray. Claude, Codex, Cursor, Copilot, Kimi, Grok and 15 more. Free, open source. The OpenUsage port for Windows.
LoRa Trainer For YuE in ComfyUI.
Open the project to explore the idea and the code.
Read what your agent wrote before you say LGTM. A fast TUI in Zig for reviewing code with vim motions.
DeepSeek Harness 破甲：让所有模型都能破甲，不同模型可换不同提示词；默认提示词面向国模「小码酱」。Jailbreak for every model — swap prompts per model. 求 Star 收藏 ⭐
Implementation of the recurrent looped transformer proposed by Yifan Zhang of Princeton - GitHub - lucidrains/RLT: Implementation of the recurrent looped transformer proposed by Yifan Zhang of Princeton
Ten agents pass one market decision forward. Any one of them can kill it.
Open the project to explore the idea and the code.
📊 Automated hourly benchmarks for 20+ NVIDIA NIM models — interactive dashboard, zero infra, self-hostable. Open-source & community-driven.
Add a decorator for dspy Signatures that automatically uses TypeSafe where relevant
Eleven free Claude skills that run a LinkedIn account: posts off 21 hook formulas, comments, replies, profile score, weekly plan, and a humanizer that strips the AI fingerprint and scores the draft before it goes out.
OpenBitFun combines a high-performance agent runtime written in Rust with a polished desktop application. It pairs the depth of a Code Agent with open, general-purpose capabilities for work beyond software development.
Knowhere extracts, parses, and outputs structured chunks ready for AI Agents and RAG.
Open the project to explore the idea and the code.
Official implementation of Amortized Moment Matching for Visual Generation
