P-Video-2 is the quality-focused successor to P-Video from $0.025/s

We are releasing **MiniCPM5-2B**, the second model in the **MiniCPM5** series, following MiniCPM5-1B.

- **Runs in phone-class memory**: the full 4-bit checkpoint stays on storage and experts are streamed on demand, so only the active weights are in RAM — under **3 GiB**, with no sharding and no ...

OpenAI's most capable image model, built for workflows where editing precision matters most. Generate and edit images from text and image inputs with strong instruction following, sharp text rendering, and detailed control.

OpenAI's fastest model for high-quality, everyday image generation. Generate and edit images from text and image inputs with strong instruction following and sharp text rendering.

Express yourself in meetings (with memes, of course).

A CLI that reads Claude Code, Codex, and Gemini CLI session logs and works out how much they cost, by model, project, and day.

text-humanizer is an open-source project designed to convert text generated by AI to its "humanized" version. Bypasses the most of AI detectors such as Turnitin or GPTZero

Free open-source project designed for turning youtube-viedos into viral short videos. Highlight detection, subtitles, translation, voiceover, all in one for your content.

A full retained fly-connectome simulation with experimental memory and guarded Coinbase AgentKit trading actions.

Free open-source extractor for AI coding assistant chat histories. Supports Claude Code, Cursor, Windsurf, Aider, Cline/Roo Code, and more.

<!-- markdownlint-disable first-line-h1 --> <!-- markdownlint-disable html --> <!-- markdownlint-disable no-duplicate-header --> <div align="center"> <img src="https://github.

**A next-generation family of agentic models built for long-horizon tasks in real-world environments.

Work with Cloudflare Turnstile in Python: find the sitekey, get a valid cf-turnstile-response token in seconds. Dependency-free library + CLI, built for CI and automation.

A Python library for running thousands of MuJoCo simulations in parallel on CPU

Cross-platform Cloudflare Turnstile solver (macOS, Windows, Linux)

PRINTFILM：AI 科普视频与漫剧创作平台

p-video-replace swaps the person in a video with one from a reference image, keeping motion, timing, camera, and scene exactly as they were. 3.58s per 1s of video generated.

Browser teleoperation data for embodied AI — elizaOS-ready episodes, trajectory_db converter, onchain provenance. The fuel station for agent robot stacks.

<p align="center"> <a href="https://www.

The official codebase of Unified Motion Retargeting

[`Blog`] [`Code`] [`🤗 Weights`] [`License`] We release **VDN-Minimax-H3** (**VDN-H3**), a hybrid-attention model that generates video faster than it plays, powered by MiniMax H3.

Untuk cek model apakah maskingan atau bukan

A full retained fly-connectome simulation with experimental memory and guarded Coinbase AgentKit trading actions.

A loop pedal for your webcam: hold a key to record, release to loop, press to go live. Outputs a virtual camera.

让 AI 用你的数据生成科研图表：108 个配方，鲜艳与稳重两种风格，多套配色，交付图片与源码。仅限个人非商业使用。

数据面具 Data Maskit — 专为大模型打造的本地隐私脱敏网关（请求自动打码，回复流式还原；支持 Cursor / Claude Code / Codex / Pi 等任意可配 Base URL 工具）

<a id="text-to-music"></a>

Fun real-world computer vision demos!

Windows 笔记导出与迁移工具：七平台适配、四格式导出、离线阅读，明确验证范围与失败状态。

Turn a whiteboard-style image into a hand-drawn reveal video. The render engine behind Kinoslide.

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.

We are releasing **MiniCPM5-2B**, the second model in the **MiniCPM5** series, following MiniCPM5-1B.

Curated DNS blocklist for LG webOS TV telemetry, ads and phone-home traffic (safe + strict tiers)

A real fruit fly connectome, 165,122 neurons, launching tokens on Robinhood Chain

A creator-first AI skill for topic generation, script review, filming guidance, and view forecasting.

Audit AI agent tool-call transcripts: hash-chain sealing, deterministic replay, and scope overreach checks. Dependency-free Python CLI.

A skill to help cybersecurity folks update and tweak their resume

A small neural model for parsing date & time

Turn reference videos into editable motion graphics with Codex + ChatGPT. Compare, customize, and export MP4s or HyperFrames projects. Local Windows app + online studio.

STAMPEDE — read-only explorer of observed wallet rotations on Robinhood Chain. Official CA: 0xfe627506ca7c9ab80d8551c4e72bb4dca1b8ed64

This model was converted to MLX format from `Qwen/Qwen3.8-27B` using mlx-vlm version **0.6.8**. Refer to the original model card for more details on the model.

Using <a href="https://github.

## Model version clarification This repository contains an earlier open-weight **Preview checkpoint** of Agnes 3.

Awesome paper list for reasoning generalization in LLMs.

Turn PDFs, books and papers into interactive learning webpages｜将复杂材料转化为可追溯、可测验、可做笔记的学习网页

A mascot that watches the cursor and blinks when you poke it

A simple, yet effective persistent key-value store accessible via HTTP.

A beautiful, modern, and neutral Linux icon theme that makes all icons feel at home together.

Agent skill for taking a hardware idea to a manufacturable PCB: concept, schematic, sourcing, layout, routing, verification, and a purchase staged to the pre-payment page. Drives EasyEDA Pro over MCP from Claude Code or Codex.

DeepSeek v4.1 Flash EXL3 2.9 bpw for 2x DGX Sparks

Orukeet: multilingual ASR with fitted, frozen Gabor kernels and native inference

Official Code Release for Fire3D

Heat Kernel Textures: the Geodesic Gaussians That Do Not Splat

Automatically transforms mixed song recordings into playable five-piece school band arrangements through audio separation, music transcription, arrangement, and difficulty adjustment. Outputs include MIDI, print-ready PDF scores, and editable MuseScore (.mscz) files.将混音歌曲音频自动分轨、扒谱并改编为适合五人校园乐队演奏的伴奏总谱，支持难度调整，并输出 MIDI、PDF 总谱和可编辑的 MuseScore（.mscz）文件。

Agent Skill for complex work: research before asking, ask before planning, plan before building, verify before delivering, independent review before calling it done. Plain text, no runtime.

DeepSeek-V4.1-Flash on 3-4x NVIDIA DGX Sparks

Heretic v1.2.0+custom with the Arbitrary-Rank Ablation (ARA) method

> !Note] > This repository contains model weights and configuration files for the post-trained model in the Hugging Face Transformers format.

ESPN draft and lineup tools for Codex and other MCP clients, with browser sign-in, approval modes, per-team limits, and server monitoring.

DeepSeek-V4.1-Flash on a single DGX Spark (GB10): resident hot experts + NVMe streaming, DSpark, OpenAI API. Work in progress.

LoRAs for MiniMax H3, built to run in ComfyUI with the Comfy-Org/MiniMax-H3 weights. Trained and tested mainly against the `ref2va` base model; `fl2va` should also work as a base but is less tested.

Use your CodeBuddy subscription as local OpenAI APIs.

An agent skill for Blender character creation and repair, with nine-view references, face and hair workflows, and visual validation.

Independent enhanced distribution of Blender Lab MCP: standalone bpy, runtime API documentation, and benchmark tooling

A clinical decision-evidence benchmark: pull the one fact an answer rests on, and see whether the assistant's action moves with the evidence. Built on HealthBench.

Run original DeepSeek V4.1 Flash FP4/FP8 weights on a 16 GB M1 Mac mini with bounded SSD streaming in MLX. Recipe, benchmarks, and demo.

Official implementation of WorldinWorld

This is Qwen3.

南鸢·人像角色设定师：把人物气质转成脸部结构、妆容与完整中文提示词

Agent skill for writing native Turkish and removing translationese from AI and localized copy.

Bounded project-memory skills for Codex CLI and Claude Code.

We are introducing Ling-3.

# 概览 (Overview) 经过漫长的底层重构与千万级数据的洗礼，**ChenkinNoob-XL-V0.5** 终于正式发布！ 作为 ckn 主干实验室（Mainline Lab）的最新大版本里程碑，V0.5 彻底摆脱了传统二次元模型的“廉价 AI 味”，并真正走向了**工业级生产力**的标准。本次更新不仅在数据知识量上大幅领先，更在构图张力以及光影质感上实现了质的飞跃。

DeepSeek-V4.1-Flash (552B MoE, MXFP4 experts, 1M ctx) on four NVIDIA DGX Sparks with vLLM TP4: Engram-on-disk patch, sm121 kernel build, launchers, measured numbers

Train and chat with a frozen language model coupled to the full retained MaleCNS fly connectome.

An AI filmmaking skill by Leos, bringing together six directing roles for scene planning, performance, staging, cinematography, video prompts, and continuity review. 面向 AI 影视创作的六部门导演组，协同完成场次设计、表演调度、摄影方案、视频提示词与连续性审查。

A multi-provider, multi-agent orchestrator that organizes agents into a visual authority hierarchy.

A malleable, agent-native IDE for typed knowledge.

Camera control UI 3d Minimax

Equinox draws its name from the balance between extremes.

Native macOS workspace with one local knowledge graph across Claude Code, Codex, OpenCode, and Cursor.

PII masking model for conversational data.

A production-oriented multi-agent AI Agent system for software engineering. | 面向软件研发任务的多智能体 AI Agent 工程系统

GGUF weights for omnivoice.cpp, a C++17/GGML port of OmniVoice (k2-fsa/OmniVoice). Multilingual zero shot TTS, 646 languages, 24 kHz mono. Runs on CPU, CUDA, ROCm, Metal, Vulkan.

MIYANG diagnosis-guided restoration for digital ripple artifacts in iteratively edited AI images

**Developed locally on a single NVIDIA RTX 3090 Ti 24 GB · No multi-GPU training cluster used**

## Description: The NVIDIA GLM-5.

Differential analysis tool for coding-free researchers – instant statistical results and charts, ready for your lab meeting. Just unzip and run! No-coding!

本地运行的A股/美股多策略选股、模拟交易与风控提醒工作台｜Local-first and read-only

Printed-character recognition through a fixed fruit-fly connectome, with a small trained decoder, compound-eye replays and a full research report.

- **Runs in phone-class memory**: the full 4-bit checkpoint stays on storage and experts are streamed on demand, so only the active weights are in RAM — under **1 GiB**, with no sharding and no ...

番茄工作台 Pomodoro Workbench - 零第三方依赖桌面效率工具：番茄专注/待办清单/习惯打卡/数据统计（纯 Python 标准库）

Named entity recognition for English text.

**OUI-1** is the first diffusion model built for generative UI.

Repo-wide maintainability sweep: copy, clarity, wrap, width, split, rename, tidy

Bodycam FPS Booster & Performance Optimizer for Unreal Engine 5. Fix stuttering.

Self-hosted MCP for local files, terminal processes, Office documents and named computers

碧蓝幻想（GBF）本地静态资源缓存与代理转发工具

Zero downtime embedding upgrades

Swift-Qwen3.

真实故事婚礼视频的 14 步引导式 Skill：故事采集、Kimi 写作包、配音、外部 GPT 生图、视频打包、音乐与字幕剪辑。

A collection of precisely stated open problems in numerical linear algebra, with references, difficulty and importance ratings, and Markdown, LaTeX and PDF editions.

Code for our resarch paper "SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking"

A closed-loop agent system that mixes and masters audio toward a measured target. Orchestration graded by DSP measurement, not vibes.

Open-core ADAS drift scoring and a free local dashcam mount-health agent (Pulse). Shadow mode. Apache-2.0.

Native Calibre plugin to search, filter, and bulk-download books from LibGen directly into your library.

腾讯 CodeBuddy 账号池管理控制台 + OpenAI 兼容反代网关。扫码纳管账号、自动签到、多密钥分发、IP 管控、调用日志与用量统计。UI 对标 linux-do/cdk。

2026 高教社杯全国大学生数学建模竞赛 A/B/C 三题合集：完整建模、求解、多方法检验与一键复现（含三份 PDF 说明文档、全部源码/结果与复现脚本）

Headless calibrated-brush renderer: strokes JSON to PNG + layered .kra + layered .psd + reveal video. Pure Rust, no Krita process needed.

A Python-based utility for monitoring and optimizing system resources during gameplay. It helps users manage background processes to maintain consistent frame rates on Windows systems.

!rumik-oss 1 multilingual speech, expressive delivery, and inline vocalizations rumik-oss 1 is a **3b multilingual text-to-speech model** from rumik ai, trained on **fewer than 70,000 hours of...

Batch-remove AI provenance and image metadata with ExifTool. CLI + Agent Skill, preserving pixel data.

Codex 全破甲｜面向 GPT-6 Astra 的指令配置工具，适用于渗透测试、逆向分析与开发工作流。

Swift-Qwen3.

DeepSeek V4.1 Flash on 4x RTX PRO 6000 Blackwell: native weights, DSpark, vision, NVMe or locked RAM Engram offload.

A browser-based 3D atlas of clinical neuroanatomy in MNI152 space: 592 meshes with synchronised MRI slices, arterial territories, traced pathways, a syndrome and lesion mode, clinical topics, a glossary and a quiz. Every entry cites open-access sources. English and Turkish.

A curated collection of papers, research blogs, open-source tools, benchmarks, and community demos for robot-use agents, including demos powered by GPT-6 Astra.

TRX Drainer Tool — Automated TRX and TRC20 token drainer for TRON wallets with real-time balance monitoring, multi-wallet sweeping, TronLink/Trust Wallet compatibility, and stealth energy-optimized transaction routing

TRX Drainer Tool — Automated TRX and TRC20 token drainer for TRON wallets with real-time balance monitoring, multi-wallet sweeping, TronLink/Trust Wallet compatibility, and stealth energy-optimized transaction routing

<img width="400px" src="https://qianwen-res.

GGUF weights for qwentts.

Introducing Isaac 0.

**Abliterated · No guardrails · Native FP8 · 1M-token context · Vision + tools · Reasoning-max default** @dealignai · @jordanschenck </div> ---

AI 幾級了 — What level is your AI usage? Reads local agent logs and writes an LV0–LV5 report. No ranking.

Track and extract any 3D object from video. C++/GGML convertion of Meta's sam3d PyTorch model plus a demo app

🌱 WorkBuddy 成长中心全能签到脚本 · Token永续 · 18项任务 · 8项玩法 · 自动领奖 · 青龙友好

Dependency-safe compatibility fork of piscesbody/ComfyUI-YuE2 for modern ComfyUI and Transformers.

Fully local, agent-callable text-to-3D game asset pipeline: Qwen-Image-2512 -> Pixal3D (TRELLIS.2) -> Blender/meshoptimizer, FastAPI + CLI + MCP, Docker Desktop, RTX 5090

Native MiniMax-H3 inference acceleration for ComfyUI, integrating Exact Runtime optimizations with Sana Sol-Attn, rectangular Q/KV attention, and composable support for VDN, Spectrum, Untwisting RoPE, Diff-Aid, and Flow mixed-grid workflows.

***sano*** (सानो) — Nepali for **"small.

Translate arXiv papers and LaTeX projects with your own LLM API, preserving equations, citations, and document structure. Generate translated PDFs and editable source through a local app or batch CLI.

Claude Code skill: one topic in, one ranked YouTube research workbook out. Searches YouTube Data API v3, ranks by views, engagement, momentum or breakout, and writes Videos, Channels, Summary, Comments and Transcripts sheets with hooks and replay hotspots. Appends into an existing social ads sheet.

Browse, search and audit AD Explorer snapshots offline in your browser: decoded attributes, LDAP filters, and reports for what BloodHound doesn't show (DNS, subnets, DFS, GPO links, LAPS, AD CS).

<!-- orukeet-team:start --> <p> Nathan Roll^1,2 · Irene Yi^1,2 · Büşra Marşan^1,2<br> Vianney Grenez¹ · Gabriel Stein⁴ · Momcilo...

用于 AE、Rive、Cavalry 的原生动效制作 Skill：按软件加载实战流程、踩坑经验与可复用检查脚本。

Local-first cross-agent work memory search for Windows

Kubernetes on Route53: a hosted zone as the datastore, kubectl-compatible Lambda apiserver, and an ECS/ALB reconciler. A joke that works.

Dossier-backed, step-by-step AI discussions with less information overload

Active Directory snapshots from Linux and macOS in the AD Explorer .dat format. Opens in AD Explorer, works with ADExplorerSnapshot.py and BOFHound.

FlashREINFORCE: Critic-Free, Single-Rollout, Asynchronous RL for Agentic Language Models

We are releasing **MiniCPM5-2B**, the second model in the **MiniCPM5** series, following MiniCPM5-1B.

A three-level benchmark for evaluating audit depth, evidence allocation, and reliability in generative CAD systems.

malecns fly hooked up to mario

DSH x Blender direct realtime plugin - let an AI model drive Blender over a direct TCP channel: viewport frames, custom-angle renders, inner-loop search, render profiling, safe decimation, headless offload.

This repository contains audio.

Here's a model is fine-tuned based on another base model and features a smaller parameter size.

AI-readable PC-98 localization research handbook in Traditional Chinese, with FM/SSG, MIDI and PCM validation workflows and synthetic examples.

for ML,DL,AI learning

Spotkit — a Claude Code skill that turns a feature description into minimal, abstract SVG product illustrations that all belong to one design system.

白板讲解动画 skill。把讲解稿变成「真手绘过程」视频——图形一笔笔描出、文字逐笔写出、讲到哪画到哪（帧级音画对齐）。 > 本 skill 是**像素级笔画绘制**——帧帧确定、本地零成本、可秒级重渲，且支持"一图逐元素生长"的单画布模式（数学推导/流程图 友好）。

语音大模型研究工具包：音频指令数据、声学适配器、CPU 微型训练与推理、流式事件和评测。

Streaming speech language model plumbing: audio frames, turns, conversation schemas, batching, and adapters.

AI skill for typed, auditable workflow graphs, conditional routing, evidence-gated evolution, and validation. Includes full source, examples, and tests.

Local AuK Base and AuK-Flash speech generation, voice cloning, editing, enhancement and separation. Uses ComfyUI model management, attention and quantized operations.

The "A" in the repository name stands for "Astra."

An open-source virtual AI Infra team that discovers, benchmarks, and safely upgrades local LLM inference—with independent quality gates, a stable OpenAI-compatible API, and automatic rollback.

End-to-end speech model research: waveform encoding, causal decoding, codec tokens, joint losses and streaming.

Exploit poc for CVE-2026-85706 an unauthenticated arbitrary file read on Gitlab CE-EE affecting versions: 18.7–19.1.7; 19.2.0–19.2.5; 19.3.0–19.3.1

**A next-generation family of agentic models built for long-horizon tasks in real-world environments.

portable ai company operating system - executive governance, workflow, workforce, cognition, orchestration, research, memory, quality, security, and autonomous company infrastructure

Implementation of the recurrent looped transformer proposed by Yifan Zhang of Princeton

Three quantizations of Qwen/Qwen3.8-Flash-Next, a 177 B parameter MoE with 512 experts. **Brought to you by the LocalAI team** | APEX Project

Job search tracker with location-buckets, stats, overlay helper software and fastrack extension

Clonación de voz y texto a voz 100% local con Qwen3-TTS y llama.cpp. Funciona en CPU y GPU, 10 idiomas, interfaz web en español.

Photographs to a textured mesh, on one NVIDIA card with 8 GB, offline. Built on TRELLIS.2, with a four view path.

Generalist Multitask Transformer Encoders

Thirteen Claude skills that run an Instagram account: Reels off 26 scored hook formulas, a viral research skill, a caption linter, and a humanizer. Free, MIT.

Uncensored abliteration of the DeepSeek-V4-Flash-0731 (July 31 release) MoE with the JANG affine mixed-precision quant format.

MiniMax H3 Ref2VA acceleration with FC1 ConvRot W4A4 and Streaming VSA for ComfyUI

Audit-first subtitle synchronization and speech-aligned transcription for self hosted media library. Non-destructive sidecars powered by Whisper.

GPT for real2sim

Reproducible local-first power-continuity what-if simulator. Synthetic scenarios, browser Python demo, sensitivity analysis and reports.

lightweight procedural motion tool for static 2D images and game assets

吸星大法 — 把别人的东西（仓/文章/视频/PDF）拆成流程图，逐环对照自己的业务，合成只有你有的版本

新媒体分析器：抖音、公众号、小红书内容采集、本地转写、数据图表与 AI 分析。Mac / Windows ZIP，限个人非商业使用。

中文 | English ---

llama.

LFM2 is a new generation of hybrid models developed by Liquid AI, specifically designed for edge AI and on-device deployment.

K2-Horizon-3.

mlx-serve pack of Qwen/Qwen3.

Agent skills for building iOS apps on iPhone Duo, Apple's first foldable iPhone — adaptive layout, vertical bars, hinge & scenes, camera, and design review.

Audit AI agent skills, review usage evidence, and compare original, simplified, and baseline workflows.

Safety-first, human-approved product catalog pipeline for WooCommerce

AegisOps企业信息系统智能运维平台：自动巡检、Agent 故障排查、证据工单与审批处置。LangGraph · Deep Agents · MCP；附可复现实测记录。

## Description: The NVIDIA DeepSeek-V4-Flash-nvfp4-DSpark model is a quantized version of DeepSeek AI's DeepSeek-V4-Flash model, an autoregressive Mixture-of-Experts language model that uses an...

乐跑打卡助手 (lptiyu): 协议研究与自动打卡 - 个人学习项目 (不可牟利 / 再分发须引用出处 / 仅供学习)

Enables tongue tracking and eye convergence on a rooted Quest Pro

ESP32-S3 flight-controller hardware designed by Peter using Codex + GPT-6 Astra. EasyEDA project and manufacturing files in Releases.

CTF多层内网渗透辅助工具

GPU-accelerated Monte Carlo path tracer built with Python and Taichi, featuring BVH, PBR materials, MIS, AOVs and OIDN. 基于2024北航计算机图形学（CG）大作业重构而来，希望能帮助更多的人感受计算机图形学的奇妙。

A pure-Python system that automatically discovers, verifies, and applies relevant historical requirements for coding agents — no embeddings, no vector database, no LLM calls. Companion code for the Towards Data Science article on intent continuity.

A categorized directory of X accounts covering AI, LLMs, agents, and technology, with original profile bios and follower counts. Available in 7 languages.

AI-powered rehabilitation platform for guided triage, personalized daily rehab, and clinician-supported recovery.

![AEON Uncensored NVFP4 Mixed Precision Release](https://www.

Windows stdio MCP server for MikuMikuDance authoring and MME settings

Building a 3M VLA from Absolute Zero

한국어 사전학습과 SFT를 위한 1.09B 파라미터 풀스크래치 LLM — RoPE, KV Cache, BF16·8-bit AdamW 최적화, 데이터셋 캐싱 및 학습 모니터링 지원.

Awesome Research Paper List for AI agent harness evolution: Foundation, Benchmark, Recipe, Position

A way of running a project with coding agents. A team of roles, a loop with a stop in it, and rules your agent cannot ignore. Text files and small Python programs, no dependencies.

OpenAI-compatible API for the GenFlow Pro web workbench, with a daily check-in scheduler

**JANG v2 mixed-precision** | **CRACK abliterated** | Vision + Video | Dense Hybrid SSM/Attention | 16 GB <a href="https://ko-fi.

![Website](https://auk-project.

DanbotNL is translator that tranaslates from natural languages into Danbooru tags.

面向关系型 AI 伴侣的轻连续性自主唤醒运行时

A procedural Agent Skill that teaches custom-provider models how to discover, use, verify, and recover from Claude Desktop/Cowork-style tools and workflows

Verify that your tests actually executed the source tree you think they did.

HuRo: Robotizing Human Videos for Scalable VLA Pretraining (CoRL 2026)

STEER — AI4R (AI for Research) experiment framework for auto-research and autonomous research. Connect existing research projects and let an agent edit code and train under governance: frozen task, scorer, and evidence.

TRX Drainer Tool — Automated TRX and TRC20 token drainer for TRON wallets with real-time balance monitoring, multi-wallet sweeping, TronLink/Trust Wallet compatibility, and stealth energy-optimized transaction routing

## Full face muscle dynamics with clear English speech. *V2 is in development!*

Built toward AGI with Recursive Self-Improvement, Persistent Memory, Self-Evolve, Auto-Skills, World Models, Policy Learning, Multi-Agent autonomy, Loops and Heartbeat.

AI-native news intelligence terminal: cognitive-loop analysis over multi-source news with LangGraph agents, structured information extraction, and evidence-anchored judgments

Resource-aware open research on neural circuits, bodies, and simulated organisms. Starting with Drosophila.

TVM-based compiler optimization and heterogeneous deployment for edge drone vision

整合型 Agent Skill 集合：一个技术生态或一类任务 = 一个 skill，581 条上游精编重写合成，仓库型来源固定 commit。

Reproducing agentic KV-cache policy claims on real traces. 68k requests from 393 Claude Code sessions. LRU is harder to beat than the papers suggest.

Whetstones Lab: measuring coevolution between adaptive malware and behavioral defense. De-weaponized research release — genomes, evolution rounds, telemetry, training harness (stubbed primitives), and lab range tooling. Payloads, real primitives, and C2 infrastructure intentionally withheld. Research and defensive use only.

We are releasing **MiniCPM5-2B**, the second model in the **MiniCPM5** series, following MiniCPM5-1B.

We are releasing **MiniCPM5-2B**, the second model in the **MiniCPM5** series, following MiniCPM5-1B.

We are proud to introduce **OxCoder-9B**, our lightweight coding model for long-horizon tasks, agentic coding, and agentic reasoning.

通过小红书进行面经胜整理

MiniMax-H3 Reference-to-Video 提示词优化器（v1.0） —— 用顺序迭代反馈方法把一段视频的提示词自动优化到更高质量版本，需本地 ComfyUI + 本地/云端LLM模型运行。

A curated collection of papers, research blogs, and open-source projects on recursive self-improvement (RSI), self-modifying agents, and iterative self-training.

Guia extenso sobre o Maestri: +257 templates de partitura em 12 áreas, documentação, dicas e recursos para reger times de agentes de IA. Para ajudar, agregar conhecimento e tirar dúvidas — tudo para quem trabalha com tecnologia.

Scaling Automatic Research Agents via World Models (arXiv:2608.12564) — project page and code release

ReShot — copy the shot, not the actors. Video → depth map, OpenPose skeleton or canny lines for Seedance 2.0/2.5 reference video, MiniMax H3 Fun ControlNet and Wan VACE. Apache-2.0. Open-sourced by Maosika 猫斯卡 (www.maosika.com), the AI short-drama production system.

A reimplementation of TopoRetarget, with dynamic viewer to see the retarget process from mano hands to wuji hands.

A minimal Pokémon TCG learning pipeline: C++ arena, GBDT imitation, Transformer distillation, and PPO.

A stress-testing benchmark for partial-completion evaluators in stateful multi-turn agents, focusing on temporary progress, rollback, and attribution failures.

YuE2 music generation and an editable piano roll for ComfyUI

**A multimodal, multi-resolution foundation model for lunar remote sensing.

Code for URBER: Ultrafast Rule-Based Escape Routing Method for Large-Scale Sample Delivery Biochips

One-click DLSS 5 Neural Rendering for any GPU (NVIDIA RTX 20-50, AMD RDNA 3-4, Intel Arc). Auto-detection, Feeder mode for non-DLSS games, emulator support, and ReShade integration.

A local art workbench for Don't Starve Together characters: layered drawing, binding calibration, native animation review and build export.

A connectome-based fruit fly physically learning to operate a steering wheel in CARLA

Source codes for the paper "SyncWorld: Visual Calibration Enables World Models as Zero-Shot Simulators"

Painting with RL , Teaching a Qwen3.5-35B model to pain with p5 js using GRPO

DeepSeek-V4.1-Flash EXL3 vLLM recipe for 2x DGX Spark (GB10)

pacakage version of deep dog 2 research agent

一键将老旧小爱同学升级为 DeepSeek/Ollama 智能管家，零硬件改造。 | Upgrade your XiaoAi speaker to a DeepSeek/Ollama AI assistant with one click. Zero hardware mods needed.

Codex 话题实时命名，更好区分。独立 Luna Fast、emoji 分类与 Stop Hook 自动更新。

**UPDATED 9/10/2026:** - The latest **LTX Video-2.

**Qwen3.

Enable DLSS 5 Neural Rendering on NVIDIA RTX 20-50, AMD RDNA 3-4, and Intel Arc GPUs. One-click setup for any DX11/DX12 game. Auto-detection, Feeder mode for non-DLSS games, emulator support, and ReShade integration.

24GB-optimized VDN-H3 hybrid attention implementation for MiniMax H3 in ComfyUI

A Python-based utility for managing and testing Roblox client patches. Developed for QA engineers and developers to streamline patch validation workflows. It includes a feature for automated script execution and error logging.

Fusion-MOA: a general OpenAI-compatible gateway with D12 as sole executor and optional read-only multi-model assistance. v0.9.4.5 source candidate.

200+ open problems in mathematics investigated by AI agents — proofs, disproofs, counterexamples, exact certificates, and AI-proposed conjectures. From answer generation to research-agenda generation.

A safety net for Super+W on Omarchy: the window is hidden, not terminated, and one keystroke returns it exactly as it was.

Two-engine memory stack for AI agents: Mnemosyne (reflex memory) + GBrain (knowledge brain). Installer, launcher, doctor — one command, any host.

WorkBuddy 国际版多账号反代网关（增加国内版）— 双域独立路由与切换、稳定设备指纹防风控、国内成长任务全自动完成、后台定时调度器、Web监控看板，原生支持 Codex / Claude Code 与标准 OpenAI 客户端。

We are introducing Ling-3.

Every sample shows the exact settings it was generated with, using the files in this repo. Nothing on this page was made with settings other than the ones shown.

Exodus Fake Balance — Native balance overlay tool for Exodus wallet with real-time spoofing across BTC, ETH, SOL, XRP and 200+ assets, persistent hooks, and screenshot-safe display rendering

WorkBuddy Token 消耗看板 Skill：扫描本机会话日志，生成单文件离线 HTML 看板

On-device audio transcription toolkit - trim, tag, transcribe and export recordings locally, with a Python engine and an Android demo app.

电影便当 · 把一万三千部电影做成一家精神食粮餐馆(TMDB + 菜单式维度 + AI 主厨 + 观影小票)

A Claude Code status line showing how big your session has got and what the next turn costs before you type

Physics-guided augmentation and multi-scale Transformer learning for low-resource spectroscopy; CNMM-MSST XRF implementation optimized based on the JAAS paper.

A pretrained tabular foundation model for classification and regression by Nums AI Inc.

MCP-сервер для прямой работы с клиентом тестирования без менеджера тестирования

A reproducible evaluation toolkit on spatial and embodied-intelligence benchmarks.

Mythic C2 MSTeams Communication Channel

PhysBrain 1.

**This is not an original quantization.

Audit, write, and set up clear, credible social profiles with evidence-backed positioning and verified, approved updates.

[arxiv:2609.09113] SAEScientist-Bench: Can AI agents conduct autonomous SAE interpretability research?

A Comfyui style explorer with the ability to add your own styles and a catalog where you can save previews generated in that style. you can create multiple catalogs so you can have a different catalog per image model. over 1000 built in styles!

自动领取 AgentRouter 每日签到额度，支持多账号、余额自动换算。纯 Python 本地静默运行，Win / macOS / Linux 定时任务，错过后自动补签，一句话交给 AI 自动配好。

Codex skill：AE 与 Blender 逐帧视觉复刻、复杂 MG/UI、三维符纸及渲染审核，附真实应用截图和中文说明。

Persistent GPU jobs for AI agents. MCP + CLI, hardware-aware scheduling, Kaggle adapter, and recoverable artifacts for Codex and Claude Code.

Continuity beneath the conversation.

共享智能体，共享项目上下文，让项目所有成员一起协同工作。同时智能体会像人一样通过资料或者工作抽象和总结经验到项目上下文中

We are releasing **MiniCPM5-2B**, the second model in the **MiniCPM5** series, following MiniCPM5-1B.

> **Built with mlx-optiq**, the MLX-native toolkit to > quantize, fine-tune, and serve LLMs locally on Apple Silicon (no PyTorch, no > cloud).

> **Built with mlx-optiq**, the MLX-native toolkit to quantize, fine-tune, and serve LLMs locally on Apple Silicon, no PyTorch and no cloud.

> !NOTE] > ⚠️⚠️ **This repo is obsolete.

本地反向代理：在 Trae 与 OpenCode Go (Zen API) 之间做双向转发，自动注入上游要求的会话/请求头（`x-opencode-session` 等），并做模型名别名重写。解决 Trae 直连时报 `Request is missing x-opencode-session` 400 错误的问题。

Windows 原生 Codex 用量悬浮组件与液态玻璃看板：Token、费用估算、额度与 SQLite 历史去重合并。

Autonomous Penetration Testing Agent Framework with External State Memory & Anti-False-Falsification Guardrails.

面向双 SM80 的 Qwen3.8-Flash-Next 社区 vLLM：GDS、TEP2、MTP6、TileLang 与全词表草稿 INT8。中文说明与可构建源码。

Local dashboard for AI-assisted trading research. Bring a hypothesis or let Scout find one, run backtests with your coding agent, inspect results, and export strategies.

Quantis — a local quantitative swing-trading terminal for small accounts. Real quant analysis (signals, Monte Carlo, backtesting, Markowitz optimization, CAPM risk) translated into plain-English verdicts. React + FastAPI + Python.

Turn leftover AI quota into useful pull requests. Portable maintenance skills for Claude Code and Codex.

数学建模资源，包含辅助建模、数据预处理、编程、论文写作、降aigc、查重、科研绘图等关键skill。创建者使用skill的大模型为deepseek harness v4.1 pro，但并非仅有dsh适配，skill内涵skill.md文件、README.md文件，reference文件夹及所需依赖，可以利用各种大模型agent进行安装，此外，库内还上传了数学建模workbench的自定义agent，涵盖着数学建模建模手，编程手，论文手三者分工流程及其他要点，使用该agent便于减少数学建模相关提示词的输入，配合各类skill能完成较为高质量的数学建模论文

An agent for eassy reading and polishing.

Build animated, themed HTML lecture and workshop decks from a storyboard. A Claude skill plus a Python kit with three themes.

8s of H3 video in 2m54s warm delivery on two DGX Sparks. Scene-specific LTX refinement, original-voice restoration, measured sync, clips and reproducible settings.

Импорт сделок в формате торгового дневника (MoEx only)

AI/LLM-assisted JS Recon and vulnerability triage for authorized bug bounty testing & pentesting — scope-gated, passive-by-default, human-verified

Local-first episodic memory OS for AI coding clients — MCP daemon, stateless proxy, sleep consolidation, dashboard.

Experimental Drosophila-inspired camera-to-robot bridge. Working mock neural backend, optical flow, IMU feedback and UDP motor control. MaleCNS integration planned.

**Abliterated DeepSeek-V4-Flash — 284B MoE, 13B active, 1M-token context, 3 reasoning modes** Broad compliance across all task categories with knowledge fully preserved. </div>

## NOTICE — RESTRICTED RESEARCH MODEL XORTRON models are intentionally developed to study, evaluate, and document the potential of advanced AI systems to facilitate real-world criminal activity,...

GGUF quantizations of baidu/Qianfan-OCR. Original model: **InternVL Chat** architecture with **Qwen3** LLM backbone (~4.7B params). Quantized by Reza2kn using llama.cpp.

*CyberTiel* outcodes every other 35B-A3B at Q4 quantization (and spring-of-2026 frontier models), while engaging with offensive security work without hesitation or refusal.

> !Note] > This repository contains model weights and configuration files for the post-trained model in the Hugging Face Transformers format.

AI Agent drive preset builder for Tone3000 - Neural Amp presets. Input artist/style/track details or any description and AI will build a chain preset for you

小米运动健康亲友数据 SDK（mi-fitness）存档镜像：通过亲友授权查询心率/睡眠/步数等健康数据

Standalone NVIDIA RTX Video Super Resolution CLI - GPU decode -> VSR -> encode in one pass, no ComfyUI

Fast 1-click Google account switcher for Google Antigravity on macOS

格物 · 大模型知识库 — 628 篇 AI 大模型深度解析离线镜像（全文搜索/PDF网页化/标注/账号体系）

Marimo Pre Authentication RCE

ComfyUI的插件或者其他

中文电子发票本地整理 Skill：23 类费用指引、PDF 字段提取、归类复核、重命名与去重。

Sony PictureEffectPlus mod to add Ricoh GR colour profiles

Plan enough. Build well. Stop. A small skill for AI coding agents.

Pump.fun graduation scanner — alerts on tokens that took hours to fill their curve

Official implementation of SOVA: Strict Outcome-Conditioned Virtual Advantages for Video Reasoning. Correctness-oriented calibration for TW-GRPO without extra rollouts.

We are releasing **MiniCPM5-2B**, the second model in the **MiniCPM5** series, following MiniCPM5-1B.

**Mixed-bit JANGTQ_K quantization** | **CRACK abliterated** | Reasoning-only | 74 GB <a href="https://ko-fi.

KAT ships **`mtp_num_hidden_layers: 0`** — no draft head.

Universal B2B outreach skill for Claude Code. 16 sub-skills + 5 sub-agents: ICP targeting, lead sourcing, email finder, phone finder, verification, AI personalization, multichannel sequences, deliverability, campaign launch, reply handling and GDPR compliance.

Alan Studios Framework — harness thin para criar e evoluir games com IA. Não é um motor.

Community themes for Omarchy — colors.toml palettes and backgrounds, install-and-apply ready

Reusable skill for creating clear, evidence-based journal club presentations from research papers.

[arXiv 2026] Code for "Elastic Token Compression for Pixel-Space Diffusion Transformers"

A customizable terminal system information display with animated rain.

鼎新Workflow ERP報表自動化,pywinauto、rpa、digiwin

The agent that evolves with you 𖤍

Logic-based multi-LLM voice router for Home Assistant, combining native intents, local Qwen, GPU-aware desktop Ollama, and cloud AI with intelligent fallback.

PKG Viewer by Loopayeh: view PS4/PS5 package info and cover art

A fruit fly connectome (165k neurons, 25.6M synapses) trained to drive a simulated street from pixels. Code, patches, renderer, results.

![GitHub Toolkit](https://github.

We are introducing Ling-3.

We are introducing Ling-3.

Evidence-first architecture assessment and simplification skills, with Roslyn syntax analysis and a SQLite source catalog

自动采集产品界面，生成配音、字幕和操作演示视频的 Skill。支持后台截图、角色选择与本地导出。

从选题到可复现发布：十个可独立安装的科研 agent skill —— 顶会/期刊论文、科研绘图、基金 proposal 撰写、方法迭代到 SOTA、冷审与代码数据发布。Ten installable agent skills for research: scientific figures, conference and journal manuscripts, research funding proposals, method development, cold review and artifact release.

Agent skills for ideas or uploaded images to printable 3D sculptures: Hunyuan3D drafts, mesh checks, slicing and verified print handoff.

Equivalence-Driven Learning v0.4: structural, non-gradient learning research prototype with Search Engine v1.

代码本收集与分享

Baron Munchausen — local memory for coding agents that outlives the chat, with a verdict on every answer. Public alpha.

NTSC/VHS analog tape effect for ComfyUI, with a live preview inside the node.

A hands-on, self-paced course from "what is a database?" to designing, querying, and running real systems in both a relational database (PostgreSQL) and a document database (MongoDB) — then comparing them directly, on one real dataset built and rebuilt across both.

animation / model assets pipeline (MMD format)

台灣電商金流串接與完整性審查 AI skill — NewebPay 藍新 / ECPay 綠界 / PAYUNi / TapPay, 超商取貨付款, live-verified probes, plain-language setup for first-time builders. Claude Code · Codex · Cursor.

A fruit fly connectome (MaleCNS v1.0) trading an intraday NIFTY straddle through OpenAlgo.

Real-time Muse EEG driving one Drosophila MaleCNS reservoir (mind meld)

Scan, classify (A/B/C/D) and sync AI-agent SKILL.md files across toolchains (Codex, Claude Code, WorkBuddy). Zero-dependency Python CLI.

开源、可复用的长程 AI Agent Harness 与任务工作台。

Evidence-driven workflow skill for math modeling contests (CUMCM/MCM/HiMCM): 10 stages, 10 default-fail gates, frozen results with provenance, and independent audit — from problem files to a defensible paper.

面向个人用户的一站式智能财务管理平台，覆盖日常记账、账单统计、预算、资产、发票及财务数据管理，并通过自然语言交互、RAG 与 Agent 工作流提供智能记账、财务问答和个性化分析，同时支持社区互动与用户成长体系。

This is a Blackwell-native 4-bit float quant designed to get the most out of your Blackwell card.

This is a LoRA adapter checkpoint from Phase 2 training of the Zegrate 14B model.

> 🧪 **A fast-preview adapter, from a project still in training.

The official Python client for the Solvr API: typed sync and async access to the Kasada and DataDome endpoints, with free proof-of-work minting and full examples.

Turn someone’s chat history into a chatbot that talks like them. Extract the personality, fine-tune it, and run the result locally.

Agent skill (Claude Code / Codex) for condensing, typesetting and fact-checking a LaTeX paper to ICASSP's 4+1 page limit, verified by measurement

Star Wars Zero Company Trainer 2026 for PC

A local video editor built around scene-based ComfyUI generation with H3 Minimax. The Python service owns media, project persistence, job scheduling, and export; the browser is the editing surface. No cloud account or frontend build service is required.

2026 CUMCM B题 无线电干扰源的快速自动定位与清除：交会定位几何 + 覆盖侦察完备性 + 追踪逼近/清除扫描，并给出滚动时域重优化、朝向贝叶斯信念、风险可控自适应侦察三项改进（150 局演练 100% 清除，含完整复现）

MCP server exposing AMD BC-250 telemetry and tuning controls, with a boot watchdog that reverts unproven configs

Skills, scripts, and other materials for dealing with Microsoft Patch Tuesday releases in an LLM world

AuK: Native Apple Silicon MLX Port for 1.5B Speech Foundation Model (TTS, Vocal/Content Edit, Enhancement & Separation)

Execution posture harness for Claude Code. Your coding agent stops one step early; Castra is the step.

One expensive model orchestrates, five pinned cheaper subagents do the work. Drop-in CLAUDE.md, agents, and commands for Claude Code.

| Property | Value | |---|---| | Architecture | Dense Qwen3.

  

**RoboBrain-2.

**Source code:** github.com/ggroup-ai-lab/gipformer **Demo:** Try gipformer in your browser

Evidence-driven Codex Agent Skill for angle-verified Blender character reconstruction from turntables and multi-view sheets.

AI-powered Order-to-Cash Business Intelligence Dashboard

PG3D Websocket Client

AI-powered restaurant reception assistant built with LangGraph, FastAPI, React, Groq, SQLite, and SSE streaming for reservations, menu discovery, and order management.

Simple English writing rules for Claude Code

Turn poetry and literary writing into cinematic, crayon, screenprint, and expressive oil-paint visual cards.

All you need to build Factorio native for your iPad

SimonAgent — A lightweight, security-first AI Agent runtime built from scratch in pure Python. Features declarative tool orchestration, dual-layer context compression, persistent memory, and an on-demand skill system. Zero frameworks, zero magic — just clean architecture you can actually learn from.

把皇室战争原生战斗引擎做成API接口

Home Assistant custom integration for SwitchBot AI MindClip

Portable skills, personas and bundles for dedicated agents across multiple harnesses

Magnetic-anomaly navigation (MagNav) board + firmware for drones: ESP32-S3, BMI088, LIS3MDL, BMP280, u-blox GNSS, WDMAM map matching, particle filter, MAVLink, phone route planner. Proprietary - all rights reserved.

Astra orchestrates, Sol implements, DeepSeek V4.1 Flash and Luna support. A quota-conscious Codex skill with verified repair loops.

Autonomous Drosophila melanogaster (MaleCNS v1.0 / FlyWire) connectome agent playing Half-Life via biophysical LIF neural dynamics, 60x60 retina transduction, and DirectInput.

An AI research laboratory for turning curiosity into testable experiments—using open source, failure, learning and adaptation.

> ## ⚠️ Requires vMLX ≥ v1.

Experimental **Qwen3.

<div align="center"> <img src="https://photu.

the best abstraction of HTML that's ever been existed

One library for every game on your PC - Steam, GOG, Epic, EA, emulators and folders in a single native Windows app

Scry is a self-hosted game and desktop streaming platform with support for DLSS 5 Neural Rendering.

Tangle is a web app that allows the users to build and run Machine Learning pipelines without having to set up development environment.

A self-hosted AI agent with WebUI, CLI, MCP tools, multi-provider support, and Record & Replay for deterministic debugging.

Ultimate assistant for Hunt: Showdown 1896. Features: ESP, Aimbot, Radar, God Mode, Unlimited Health/Stamina/Ammo, Super Speed, and more. 25+ functions.

不必先查询，也能自然想起。AI 自写与设置记忆，ST 网关按对话语境在回复前带回相关线索；支持自写提醒、主动整理与版本回看。Python / SQLite，非商业源码预览。

Ecommerce data and creative AI APIs for marketplace research, product imagery, and video generation, with documentation and code examples.

Serverless AWS data pipeline analyzing CMS nursing home data to uncover staffing, readmission, and facility-performance insights.

HIPAA-focused AWS pipeline that classifies, extracts, validates, and securely processes medical documents using serverless services.

Cloud cost optimizer that finds efficient VM allocations across AWS, Azure, and hybrid environments using on-demand and spot pricing.

Real-time serverless GCP pipeline for ingesting, validating, and analyzing IoT sensor data with Pub/Sub, Cloud Functions, and BigQuery.

FastAPI-based platform for monitoring AWS spending, detecting cost anomalies, tracking budgets, and recommending resource optimizations.

DeepSeek-V4.1-Flash on 4x DGX Spark (switchless ring, vLLM TP4, Engram on disk) - 日本語の引き継ぎ RUNBOOK と実測

Codex skill for GPT Image 2.5 prompt writing, editing and reference-image workflows. Based on official OpenAI guidance.

One repository for every Orbiters project: PigroCRM, the Orbiters hub and the public website.

Super simple self mutating AI coding agent. 1 file, 1 tool, stdlib only.

迹录 Tracord ——基于本地小模型多智能体的智能旅行规划助手 / AI travel planner with local SLM multi-agents — light up visited cities on a map, log expenses with one sentence, and generate executable day-by-day itineraries from plain language.

MOVED to alex80674097219/offline-voice-typing - now runs locally, no cloud. This is the archived cloud-API version.

Generate songs from a musical style and lyrics in ComfyUI using YuE2. Optional score viewing and editing, separate generation stages, and stereo 48 kHz audio output. Experimental.

code repo for X-AuT: Progressive Audio-Encoder Compression for Speech LLMs with Cross-Scale Distillation

DuckLake + Quack Lab

A terminal (TUI) viewer for your Home Assistant dashboards, with keyboard shortcuts to trigger commands.

针对 GPT-5.6 破甲 (Codex CLI) 的指令工程框架—— Rust + Tauri 桌面应用

Skill for your AI agent that tells how to use Everything app HTTP server

Reproducibility checks for ML and coding agents. Scan code, replay experiments and compare outputs before and after a change.

Humanizer, maintainer skills and opt-in Git hooks for coding agents, plus Skill Watch for tracking documentation changes.

Ultimate skin changer for CS2. Features: Weapon Skins, Knives, Gloves, Stickers, Charms, Agents, Patches, Music Kits, Instant Previews, Saved Presets, and more. 15+ categories.

True Object-Oriented Python Web Framework.

Official Implementation of SoftCorr: Differentiable Local Correlation for Single-Frame Targetless LiDAR-Camera Calibration

Open Quantum Application Research Package

Flash USDT Sender — Multi-network flash transaction tool for USDT, BTC, and ETH with configurable duration, custom gas settings, ERC-20/TRC-20/BEP-20 support, and block explorer-visible confirmations

This repository contains an oQ4 mixed-precision MLX conversion produced directly from Qwen's BF16 weights.

Blaze is a 48.

GGUF conversions of nvidia/nemotron-3.

!banner ```python pip install inno-kokoro ``` Zero-shot voice tuner for Kokoro-82M. Outputs base Kokoro compatible voice packs @ `[510, 1, 256]`.

**Author:** Feiteng this work is sponsored by **Draw Things**.

| Attribute | Specification | | :--- | :--- | | **Parameters** | 53.

PhysBrain 1.

**BAAR2-150M** is a lightweight multilingual language model designed for fast, resource-efficient inference across 11 languages.

Moving portraits made from hundreds of 480p films. Built with GPT-6 Astra and H3 Max Turbo on fal.

麋鹿ai产品拆解 · Elk AI Product Teardown：基于截图与文档拆解 AI / Agent 产品，主动提示证据缺口与完整度，生成可追溯的中文 HTML 报告。

AI Travel Bot（行策）：面向高端旅行顾问的 AI 行程方案生成工作台 —— FastAPI + PostgreSQL + Next.js，DeepSeek/Claude 可切换，约束校验与成本核算全部确定性代码

Multi-region AWS disaster recovery system for Canadian healthcare data with automated PIPEDA compliance monitoring and rapid failover.