源界:面向 Python Agent 与静态 MCP 配置的本地开源项目安全审查工具
Automatically resume the exact same conversation in the Codex desktop app on Windows after usage-limit resets, rate limits, or temporary failures.
Code for 'Reasoning on the Manifold: Bidirectional Consistency for Self-Verification in Diffusion Language Models' (ICML 2026)
QD-Rec supports decision-making through open-ended recommendation using foundation models and quality-diversity search.
Free, open-source voice typing for Windows and Android. Talk and the text is typed into any app. Speech-to-text runs offline on your device with Whisper, no cloud, no account.
外接盒子AI推荐打法从而高胜率支持弃牌术、海盗瞎等卡组的Hearthstone脚本(炉石脚本胜率63.8%,稳定上传说)
Lumen: a tvOS-style, GPU-light home screen for rooted LG webOS TVs (patches on partridgeworks/webos-launcher)
SuperJLU —— 吉大人的一站式服务:课表、成绩加权分析、绩点、考试、空教室、培养方案、OA 与学院通知;Windows exe / Android apk / 本地网页,数据只存本机
Official Spytrend MCP server connection metadata, documentation, examples, and AI workflows for Meta and TikTok advertising analytics.
UI/UX and frontend design skill for AI coding agents: Codex, Claude Code, Cursor. Art direction, DESIGN.md, design tokens, motion, accessibility, and real interface validation.
—
Unofficial Online Server for Forza Horizon 2 (Xbox One build) on PC via Azimuth.
Semantic ifs from open models, on a 3090 at home. Independent; not affiliated with Jev or TypeSafe.
Simulate Before Reality.
The long-horizon computer-use harness. Run AI agents across desktop apps and the CLI for extended periods while preserving task state and making reliable progress on complex workflows. Features fresh-context execution, durable verified state, independent auditing, recoverable progress, and native Claude Code / Codex / OpenClaw integration.
AuK: An Open-Source Foundational Model for Speech Generation and Editing
The open, local take on Jev: typed decisions from an LLM, without generating a single token
Open, Jev-compatible System One decision server on DiffusionGemma
向顶会优秀论文学科研|Oral Paper Skill — 从 883 篇 Oral 论文摘要中“蒸馏”共通做法,用于论文对照改进与学习复盘 · Skill + prompt pack (CN/EN). Learn from 883 Oral paper abstracts to improve your research and writing.
A lightweight, extensible, fully offline development platform and agent/workflow toolkit for the AI era, with an intuitive TUI, easy deployment, and complete control over your data, agents, and workflows.
CarrierSIM - Установка профилей операторов для разблокирования возможностей iPhone и iPad: VoWiFi, 5G и EVS в России и Беларуси
—
DMAD: Distribution Matching as Adversarial Distillation for Fast Visual Generation
A vLLM-inspired inference engine for looped transformers, with loop-level scheduling and execution across composable stages
🚀🚀 A multimodal System One decision model that gives calibrated answers to typed questions about screens, photos, video and text in one forward pass.
A curated list of tech companies in Spain offering a €60k+ median base salary for Software Engineers with 5+ years of experience, backed by Levels.fyi data.
[NeurIPS 2026 Spotlight] MoSE3: Learning World-Space SE(3) at Every Pixel
适用于华为智慧生活设备的 Home Assistant 自定义集成。
An agent that compiles git maintenance chores into reusable programs, then decides whether to replay them by running checks rather than by guessing the task looks familiar
✈️ 自托管的多渠道机票价格监控:五平台明细采集 → 钉钉图文推送 → 达标四层强提醒。国内航线与国际航线(五洲目的地)同一套采集-判定-推送链路;数据留在本机,配置全在网页,开源免费。
—
Decky Loader plugin for Steam Deck and other SteamOS/Bazzite handhelds: download, build and install HelixSR (a DLSS-based FSR 3.1 replacement) per game, directly or through OptiScaler, and check that it's working.
Official repo of UltraTex: Unleashing 2K Multi-View Diffusion for 3D Texturing (SIGGRAPH Asia 2026)
Official codebase for Unitas: A 3D-Native World Action Model for Embodied Manipulation
EvoSCM: Scientific Belief Revision Through Causal Model Evolution and Experimentation
A 4B diffusion grounding model for fast, precise visual evidence extraction. Unifies diverse grounding tasks through parallel decoding, with optional self-speculative decoding delivering 4.51x autoregressive throughput on COCO.
Pemutar & agregator anime mandiri (Web, Windows, Linux, Android). Ringan, tanpa iklan, tanpa akun & subtitle Indonesia.
A local acoustic control system that learns user-defined labels for the surface or location of knocks, such as desk or wall. It also counts threshold-qualified knocks within each captured window. By default, a frozen Microsoft BEATs Iter3+ AS2M encoder creates 768-D audio embeddings and an incremental classifier learns the user's labels.
DeltaStreamer: Towards Human-Like Agentic Streaming Video Understanding
AFI VISION BUILDER
Local decision engine with typed, calibrated outputs (Noul/Choice/Score): cheap→strong cascade, ECE/Brier calibration, SystemOne protocol interop. Apache-2.0.
Drop in raw talking-head takes, get back a finished, publishable video. Every rule in this repo was paid for by a rejected render.
ORA: Open Robot-Use Agents
同花顺、通达信等券商协议下单api,代替miniqmt、QMT等,L2行情
受控中文技术写作技能:ASD-STE100 的中文对应物,含 stdlib-only 确定性 linter。Controlled Chinese technical writing skill for agents.
Free HTTP/SOCKS proxies with every check on the record: passed, failed or couldn't run, and when. Hourly list, --pick --verify, --diagnose, rotating proxy, Python API, MCP server.
🚄 KTX 표, 파이썬이 끊습니다 — 열차 조회 · 예매 · 결제 · 환불 · 예약대기 · 취소표 감시를 코드 몇 줄로. 코레일톡 API 비공식 Python SDK (Unofficial Korail / KTX API client)
—
A Real-time LiDAR World Model that generates, forecasts, and resimulates within a single model.
Local Laya typed decisions on Apple Core ML and Neural Engine. Validated ports, ~5 ms short decisions on M3 Max, reproducible speed and energy benchmarks.
为纯文本模型"看图“设计更好的视觉工具箱和技能,支持多图理解,图片问答,前端UI还原、GUI 自动化等,并可选无缝接入多个主流agent,直接识别粘贴图片| A vision toolkit and skill designed for text-only llms — image Q&A, long-screenshot OCR, frontend UI restoration, and GUI automation, with optional seamless integration for Codex, Claude Code, Pi, Oh My Pi, and OpenCode
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
Marigold V2: Revisiting Diffusion Transformers for Monocular Depth Estimation
Awesome skills for submitting url to free websites. Get more backlinks for your website to get more traffic.
Frontier-class open models on a free Kaggle TPU v5e-8: GLM-5.3-Flash 320B MoE (~165 tok/s, our own JAX engine) and Qwen3.8-27B bf16 (~130 tok/s), 262k context, prefix caching. Works with Claude Code, Codex, opencode and pi.
Easy to use suite for designing a wide range of protein binder modalities.
—
ScienceBuddy: Recursive-in-Recursive Self-Improvement for Interactive Scientific Agents
world.execute(me); —ascii
—