⚡ Official code for “Massive Activations in Hybrid Linear Attention Large Language Models: Pre-Attention Spikes and Inter-Spike Plateaus”.
—
单文件 Windows 网络诊断命令行工具 · 内置 23 项诊断模块
Any website, within reach of any agent. Captures a site's native WebMCP tools and generates the bridge from the live DOM when there are none, exposing both over MCP to Claude Code, Codex and any MCP client.
微信里的话不知道怎么接?macOS 悬浮聊天助手:识别消息意图与沟通风险,GPT 生成多种话术,Jev 评估候选,一键填入微信。话我帮你想,发送你来定。
—
[NeurIPS 2026] DriveHierarchy: A Benchmark for Diagnosing VLM Driving Capabilities from Open-Loop Understanding to Closed-Loop Execution
Mental-R1: Aligning LLM Reasoning for Mental Health Assessment
Scientific literature discovery and local paper linkage workspace. Just drag and drop a single paper and watch the network grow
通过 MCP 与 AI Agent 采集 X 帖子和评论,支持博主画像、选题素材研究与轻量小时监测。
SRAM-based arbitrary IQ transmission for ESP32-S3 and ESP32-S31, with selectable sample rates, QPSK/16QAM OFDM waveforms, and E310 spectrum and constellation tools.
Emulation framework for Android & iOS native libraries
py123d_garage: a starter kit for open-loop training of end-to-end driving policies
The Proof-of-Control Standard (PoC) is a catalogue of verifiable requirements for AI agent systems.
这是一套 MiniMax H3 本地部署与性能优化整理:在 4 张 22 GB RTX 2080 Ti 上,通过 Raylight FSDP/USP、双组NVLink、多卡 Qwen 文本编码节点、SM75 SageAttention 和通信流水,让MiniMax H3 的 INT8 与 BF16 权重能够在 ComfyUI 中进行实际视频生成
Run a 177B model (Qwen3.8-Flash-Next) on an 8 GB laptop GPU. Measured: 6.6 GiB VRAM, 47.8 GiB RAM, 34-35 tok/s. The n-gram table stays on disk; the experts run on CPU.
Inference Serving Framework
[ECCV 2026] GEM-4D: Geometry-Enhanced Video World Models for Robot Manipulation
—
—
水仙的Threads 爆款雷达:实时监控 Threads 上关键词相关的公开帖子,按互动量与上涨速度排出榜单;同时按账号聚合同行,看谁在持续产出爆款。零第三方依赖(纯 Python 标准库),含关注名单、CSV 导出、互动增长曲线、通道自检与自动熔断。Real-time Threads viral content radar — monitor keyword trends, rank posts by engagement and growth velocity, and track creators who consistently produce viral content.
The SSI-263 speech chip, emulated, with the original firmware of the Speak-Out, Braille Lite, Accent-mini and Accent SA driving it. NVDA add-ons.
AgentPersonaBench (APB): do persona traits steer what agents do? Code and data release.
Learn how AI Systems work by operating them
Official repository for "Uncertainty-Aware World Model for Aerial Image-Goal Navigation".
Python + TradingView scanner for Qullamaggie-style momentum setups (Breakout, Episodic Pivot, Parabolic). Personal project, not affiliated with Kris. Not financial advice.
CONTINUUM: Verifiable semantic recovery for long-running AI agents. Semantic checkpoints (not conversation dumps), an idempotent action ledger that refuses duplicate side effects, and a hash-chained tamper-evident event log, all exposed as a deny-by-default MCP server. Framework-agnostic, Python 3.11+
Official VirusTotal MCP server for AI agents: threat intelligence, file analysis, and recovery through VTAI.
Official implementation of HOT: Robot Tool Design from Scratch via Behavior-Aware Hierarchical Optimization
RGBA VAE + extract/remove LoRAs for FLUX.2 Klein: Gradio demo and ai-toolkit patch
Concept Reel · vibe知识大赏视频生成工具与 Agent Skill,支持 Canvas 动画、字幕、确定性渲染和事件音效
A skin for VLC 3 based on Juxtopposed's redesign of VLC.
—
应用软件(直播源整理) · 聚合公开直播源,只保留电视台,按实测码率排序,导出 M3U
PhaseCoder: Microphone Geometry-Agnostic Spatial Audio Understanding for Multimodal LLMs
AI原生组织(AI-Native Organization)— 免费开源书籍,持续更新
Run Jev-style typed decisions locally on your Mac with low RAM usage and fast responses
Скилл для ИИ-агентов и CLI на Python/uv: находит и убирает ИИ-стиль в русских текстах (канцелярит, кальки, штампы), проверяет сохранность правки и оценивает текст по фрагментам в духе «Антиплагиата» с калибровкой по вашим отчётам. Русская адаптация avoid-ai-writing.
Command completion menus and history-based autosuggestions for zsh, Bash and Fish on macOS and Linux
Formal verification for real codebases with Claude Code: TLA+ and Lean 4 skills that find, reproduce, fix and prove concurrency and state bugs. MIT, by Opum AI.
Unofficial save editor for Minecraft Dungeons II (Steam / Xbox app / PC Game Pass): hero stats, items, presets and sorting, with backups
Expose any AI agent worker as an OpenAI-compatible API — plug it into 9Router (or any OpenAI client) as a custom provider.
Mathematical research agents with shared, structured state for understanding the reasoning process and coordinating exploration.
免费本地合成视频 BGM:解析 Remotion / HTML 时间轴,按场景与关键时刻配乐
借脑:把团队里 AI 用得最好的那个人的 AI,借给每个人。Let coding agents on different machines start, continue and read each other's sessions (Claude Code ↔ Codex), with per-person permissions and rules. No inbound ports.
—
Progress bars that keep filling after they're printed. Live bars in Claude Code for Ghostty, Kitty and herdr.
Agentic skill to create a hill to climb for every problem
ASI — continual reinforcement learning framework (The Alberta Plan) in JAX; trains elizaOS robot policies
一个 Windows 桌面辅助工具,用于从 online-fix.me 自动检索并下载游戏的联机补丁(Online Fix),简化手动查找和下载的流程。
Audits a repository for how ready it is for AI coding agents, and returns a prioritized fix backlog with evidence for every finding.
—
AI 驱动的自动化渗透测试平台 — 黑板架构 / 思考强度 / 验证铁律
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
Codex-style Computer Use for DeepSeek Harness: the official window2 13-method desktop surface plus Chromium automation, a visible overlay with Esc interrupt, per-app approvals and bundled skills. Windows, MIT, no Codex required.
SDK-first, independently maintained SemIf fork for fast, self-hosted semantic decisions.
Official code repository for "Into the Unknown: Benchmarking AI agents on open scientific research problems"
—
Proxmox VE 8.x/9.x 自托管管理面板:虚拟机与 LXC 容器全生命周期、cloud-init 模板流水线、 监控大盘、VNC/串口控制台、多角色权限与操作审计;另含 AI 排查助手(会话式归因、要改系统的 命令逐条批准、Token 用量统计)、SSH 防暴力破解、端口异常检测与安全基线加固。 A self-hosted web panel for Proxmox VE — VMs, LXC, cloud-init, monitoring, RBAC, audit, and an AI inspection assistant.
Causal streaming odometry for calibrated multi-camera rigs, built on a frozen 3D foundation model. Training, evaluation and pretrained weights for NCLT and KITTI-360.