Argolink-focused multi-model CLI routing with an image-capable target for Nano Banana Pro.
—
Argolink-focused LLM evaluation and regression workflow for GPT-6 Sol.
Wan 2.7 Video Edit API (`wan2.7-videoedit`): model id, pricing, parameters and copy-paste examples for the OpenAI-compatible APIMart endpoint.
Gemini Nano Banana 2.1 prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`gemini-nano-banana-2.1`) and the cost per image.
Сиденс / Сиданс / Сидэнс — Seedance 1.5 Pro prompts: tested image prompt recipes with example renders, aspect ratios, the model id (`seedance-1.5-pro`) and the cost per image.
Add to steam as a non steam game. Choose 'Proton Selector' as a game's compatibility tool. Interactively change a game's proton version without restarting steam for new versions. Use something like ProtonPlus alongside this to get new versions
—
—
—
Superproject for local scientific agents, simulation-result adapters, and secure research workflows
Code for Large-Discovery-Model (LDM)
A lightweight, Git-native knowledge base skill for AI Agents — no vector DB, no chunking, no embeddings. Just TOC + line numbers + full source.
ScanSci Harness — evidence-first AI research workbench powered by Pi.
[Arxiv ] Code for "DIVE: Dynamic Iterative Visual Evidence Construction for Efficient Vision-Language Models"
Explainable AI for detecting and diagnosing dental issues from X-rays.
Disk-backed pipeline that takes AI coding agents from raw idea to shipped code — gated phases, orchestrated subagent builds, and crash-proof state, from a single milestone to a full multi-milestone program.
AI-powered real-time object detection, tracking, and video intelligence platform using YOLO and computer vision.
The official repo of DriveVLA-M0: Failure-Aware Memory Augmentation for Autonomous Driving (ACM MM2026)
AI-powered fake news detection system with multiple ML models and explainability.
Pure MLX MiniMax-H3 text-to-video-and-audio inference for Apple Silicon.
Create reference-adapted hand-drawn follow-along videos with nib-synchronized motion, TTS, subtitles, and deterministic rendering.
Macaw — on-device AI agent for macOS. 2.7B LFM2.5 fine-tune, MLX 4-bit, 97 macOS tools, no cloud.
Corporate Waters AI Agent
An RDNA4 R9700 AI Pro ComfyUI patch for MiniMax H3 to speed up video generation.
A minimal, evidence-gated harness for Codex and Claude Code: reuse what works, add only what is missing, verify before evolving.
把任意故事自动变成完整图片素材、分镜、中文视听提示词与 API 清单的 Codex Skill
Operational tooling & discipline for multi-agent AI engineering fleets — mutation-proven guards, a Textual terminal monitor, generalized skills, and specs. Every check ships with proof it can fail.
—
面向 DeepSeek Harness 的交互式 K 线分析插件,支持多市场行情、技术指标、支撑压力位、新闻与基本面分析,并在原生侧栏中直接展示。
—
—
—
本地微信消息、历史档案与可解释情报分析工作台
一个给自己的微博账号做「茧房判断」的 Agent Skill
Solvers for the Multi-Depot Heterogeneous Fleet VRP with Time Windows (ACO, ALNS, GA, SA, TS, DE, DPSO)
Detect and type PII locally with one ProgramAsWeights neural program.
Elbi gives your data agent a verified fact store.
Agent Valley — a hands-on chapter-one agent lab: catch a model lying, give it a tool with one line, and watch the state it changed survive a reload.
AI Integrations for Temporal
Ultimate assistant for FiveM. Features: Money Hack, Aimbot, ESP (Player/Item/Vehicle), Wallhack, Speed Hack, Fly/Noclip, Teleport, Lua Executor, Vehicle Spawner, God Mode, and HWID Spoofer. 25+ functions. Undetected.
AI-readable PC-98 localization research handbook in Traditional Chinese, with FM/SSG, MIDI and PCM validation workflows and synthetic examples.
—
—
—
—
—
用 TypeSafe Jev 推荐已安装 Skill / Bounded installed-skill recommendations with TypeSafe Jev. Python CLI, Codex skill, bilingual docs and live examples.
Demo: IA detecta golpe do Pix e rastreia a quadrilha. Rust + Jev vs Python + DeepSeek em tela dividida.
Typed decisions from frozen Gemma-4-12B-it: one-token option-letter readout over stock vLLM 0.30.0 (JevBench package)
Cross-platform file sync tool with U disk write-life protection. Supports Windows, macOS, and Linux.
An open-source agent runtime inspired by OpenAI’s Dots: persistent AI agents that pursue goals, react to events, and keep work moving 24×7.
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 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 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 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 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.
SearchJev: a fast and calibrated System-1 model for search agents (training code)
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 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.