[Public preview] An advanced AI Benchmark for math, produced by mathmo at St John's College, Cambridge.
Desktop GUI for Qobuz hi‑res / lossless music downloader. Built on streamrip | Electron + React + FastAPI
—
[Public preview] Hard human-written reasoning puzzles: 749 answered items in 15 topics (707 scored core), with protocol and scorer.
An open-source multi-agent platform for GitLab MR code review, governance, and CI self-healing.
PeakPulse AI — Real-Time Financial Market Anomaly Detection & Time-Series Machine Learning Engine
A Windows tray app that interrupts you when you yell. Measures your voice on first run rather than guessing a threshold.
This repository provides Unitree G1 deployment code for GRIT v0.0.1, a whole-body motion-tracking control algorithm for humanoid robots.
Local-first, evidence-aware project memory for AI coding agents.
Mini Program Engineering Skill Suite is an Agent Skill suite for evidence‑first mini‑program development. It helps agents bring WeChat or other mini‑program projects from vague intent to reliable engineering work: project intake, product specification, architecture, implementation, UI & device adaptation, debugging, verification, release readiness.
Sub-millisecond query routing & cost-aware execution for GraphRAG architectures. Adaptive-AtomRAG is a highly optimized conditional routing gateway designed to resolve the latency and cost bottlenecks of heavy Atom-Entity Graph (AEG) RAG systems.
Fast-AtomRAG-Core
seedream-5.0 API (seedream5.0 / seedream 5.0): 1K-layer $0.0146; 1K $0.0293; 2K-layer $0.0293. Model id, prompts settings, curl and Python examples over an OpenAI-compatible gatewa
Find the most stable nearby Fastly CDN route and pin it in hosts to speed up GitHub downloads and X images. Windows & Linux, menu-driven, English/中文. 自動找出附近最順的 Fastly 路線並寫入 hosts,加速 GitHub 下載與 X 圖片。支援 Windows / Linux,中英文選單操作。
Turn a GitHub repository or Hugging Face model into a finished vertical reel: video, cover art and per-platform copy.
—
One-pass anonymous Active Directory enumeration over SAMR and LSARPC — null session, no credentials, with structured reusable output.
Ubuntu on Xiaomi Pad 6 Pro (liuqin): mainline Linux integration, build tools and installation images.
rétro-ingénierie Grok Imagine 1.5 Edit Ext API: 1K $0.015. model ID `grok-imagine-1.5-edit-apimart`, OpenAI-compatible, async task + cost in every response, $1 minimum top-up. grok-imagine-1.5-edit-ext-reverse-api-fr.
—
Fetch.ai LinkedIn poster agent for Agentverse using uAgents and ASI:One
AI API gateway / relay: 307 models behind one OpenAI-compatible key. image2.5 $0.0085/image, Seedance 2.0 Mini $0.01056/sec, LLM from $0.0228/M tokens. USD settlement, pay-as-you-go, $1 minimum top-up.
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.
rekayasa balik Grok Imagine 1.5 Edit Ext API: 1K $0.015. model ID `grok-imagine-1.5-edit-apimart`, OpenAI-compatible, async task + cost in every response, $1 minimum top-up. grok-imagine-1.5-edit-ext-reverse-api-id.
AI API gateway / relay: 307 models behind one OpenAI-compatible key. image2.5 $0.0085/image, Seedance 2.0 Mini $0.01056/sec, LLM from $0.0228/M tokens. USD settlement, pay-as-you-go, $1 minimum top-up.
Modern UI restoration patch for Star Wars: KotOR — resolution-aware fonts, layouts, inventory and area map across 48 resolutions, from 4:3 to 32:9.
AI API gateway / relay: 307 models behind one OpenAI-compatible key. image2.5 $0.0085/image, Seedance 2.0 Mini $0.01056/sec, LLM from $0.0228/M tokens. USD settlement, pay-as-you-go, $1 minimum top-up.
Desktop app that automatically tests Windscribe servers, protocols and ports to find what works on your network. English, Farsi and Russian.
EX520v için tek tuş root aracı
Independent from-scratch implementation of the Kimi K3 architecture (arXiv:2607.24653v1): KDA, NoPE, latent-space MoE, and the systems co-designs. Table 1 reproduced to 0.09%.
MikroTik RouterOS 的实时流量面板 · A live dashboard for MikroTik RouterOS — flow diagram, event stream, 1-second resolution
a skill make agent know how to build a lua mod of helldivers 2
Evidence-first game operations incident investigation agent with read-only MCP tools and reproducible evaluations.
A Python framework that provides a common data model and computational abstraction layer for MRI raw data
Turn travel photos into archival photo-and-rubber-stamp field note posters.
[ACM MM 2026] Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering
alpha-hunt — an autonomous job-hunting edge: scans company career pages every 15 minutes, auto-submits ATS applications the moment it's confident, sniffs out the wider market via Adzuna, and pings you on Telegram the second something worth chasing shows up.
Scaling Automatic Research Agents via World Models (arXiv:2608.12564) — project page and code release
AI API gateway / relay: 307 models behind one OpenAI-compatible key. image2.5 $0.0085/image, Seedance 2.0 Mini $0.01056/sec, LLM from $0.0228/M tokens. USD settlement, pay-as-you-go, $1 minimum top-up.
AI Powered Texture De-lighting for 3D scanned assets.
GLM-5.3-Flash NVFP4 on two or four DGX Sparks cabled as a ring, with no switch: the RiNGSiDE serving recipe
MRI research guidance for coding agents — composable skills, established tools, and evidence-linked research memory. Install: npx skills add KeWang0622/MRFoundry
An easy‑to‑use multi‑platform info aggregator for the VTuber community.
Ubuntu Desktop for OnePlus Pad 2 Pro: kernel integration, device packages and installation tools
HBG:把 Gemini Web + Flow/Veo 图片、视频能力统一封装进 Docker 与 CLI,支持本地浏览器授权、自动下载与可选媒体后处理。
—
Your AI coding agents keep working while you sleep: drive Claude Code, Codex, Gemini, opencode & more from your phone; auto-resume past usage limits.
Swing/momentum trading master methods as skills + yfinance scanners: O'Neil CAN SLIM, Weinstein Stage, Darvas Box, Wyckoff, Minervini VCP. Verified on real data.
[Public preview] Externally scored agentic ML research benchmark: 60 tasks, real competition ground truth. Open protocol, operated evaluation.