—
Native, local-first knowledge graph engine for code and project artifacts—inspired by Graphify, built in Rust, and evolving beyond it.
Into the Unknown. —— 探索未至之境。
A Codex Skill for converting Beamer PDFs into editable PowerPoint decks with native equations.
—
Nix Flake for the Zen Browser, with automatic updates
—
AI口播智能体,实体口播作弊神器
把上課教材變成 iPad 上好讀、可手寫的深色講義|Agent Skill for Claude & Codex
Terraform module for fck-nat
Map your path from a low-priv user to roo, like BloodHound, for local Linux privesc.
计算机科学与技术专业硕士研究生入学考试(11408)复习笔记与考研经验贴
Production-oriented Kubernetes learning path: 47 lessons, 35+ labs, manifests, diagrams, cheat sheets, interview and certification prep.
Interactive visualization and audio exploration of drum eigenmodes. Created by Basel Ashraf.
Your planet. The release notes. An unofficial changelog for Earth, built from real reporting.
MerchantBench is a 365-day, order-level benchmark for evaluating the long-term coherence of LLM agents in seller-side e-commerce operations.
—
⚡ Zero-reload dynamic AI model switcher & local proxy for VSCode Copilot Chat via 9Router
Animated system fetch that stays pinned above your shell while you work. Linux, no dependencies. 🦀
RSIGame: Autonomous agentic game development with recursive self-improvement
🎨 Tiny Syntax Highlighter
Edge-cloud Multi-Agent execution framework for embodied intelligence, connecting AI agents to real devices with verifiable actuation and entropy-bounded recovery.
—
—
The database you already have.
For Educational Purposes (it's a joke)
—
A native macOS assistant that reads the current screen and browser context when you need it.
fear, uncertainty and ... terminals
把 dsh 会话变成合规、可复现、可追责的 AI 投研工作台:ai-hedge-fund 回测 + risk-gate 风控 + cron 盯盘 + 记忆复盘 | compliance-first AI investment desk
[EMNLP 2026] A multilingual benchmark for repository-level unit test generation, evaluating LLMs across 5 programming languages and real-world code under reproducible, Docker-based execution.
Ask the Web Account Manager (WAM) for Entra ID tokens
A library of experiments using frontier LLMs/VLMs for robotics tasks, currently focusing on GPT-6 Astra.
Portals marketplace gifts autobuy: collection rules, fair-value checks, sell-later exits, practice mode by default
—
推荐组件(非必须):DeepSeek Harness 运行时注入器;已随 dsh-routing-suite 单仓库化保留,本仓库继续维护/发布。
Model-driven context management (Active Context Pruning / ACP) for the DeepSeek Harness — the model decides when and what to compress. Ported from billion-context-pi (ranxianglei); acp-kernel reused verbatim. CompactionEngine backend with compress/decompress/search_context/acp_status tools.
Three liquid interactions you can grab, pull, and let snap back — anchored dropdown, morphing dropdown and speed dial on one goo engine.
DeepSeek Harness plugin for previewable cross-preset session migration. Fixed-schema handoffs preserve state, source-model intent, and unresolved images; the original session stays untouched.
The deterministic tax engine for AI agents. Highest score ever recorded on TaxCalcBench (96% exact returns)
Run any coding agent, on your own servers, from anywhere
Best ready-to-run DeepSeek Harness distribution: DSH desktop GUI, DSH terminal TUI, and harness integration.
—
High-performance Piecewise Geometric Model (PGM) learned index implementation in pure Rust with strict epsilon error guarantees
Make Faster — website performance tools and site leaderboard
A judgment-first coding agent runtime
—
A lightweight, private system monitor that lives in your macOS menu bar. CPU, memory, battery, network, disk, GPU, temperature, fans. Zero telemetry.
面向 Windows 和 AI Agent 的公众号文章排版流水线:Markdown + 本地配图 → 公众号兼容内联 HTML、发布稳定版、390px 手机预览、长截图与完整性校验报告。不改原稿,不自动发布。
—
Youjian shared focus space
Self-hosted memory service for AI assistants
A small .NET library for splitting binary data streams into complete protocol frames. It handles partial and combined packets, configurable length fields, noise recovery, XOR, and CRC16-Modbus validation.
Real-time 3D digital twin of an autonomous warehouse — 20 AMRs, deterministic simulation, explainable fleet scheduling, failure injection, VLM camera perception, AI Ops Copilot and what-if scenarios. React Three Fiber + FastAPI.
Awesome paper list for reasoning generalization in LLMs.
NVIDIA Sol-Attn for ComfyUI / Triton kernel on SM89 - SM121, with zero-copy MiniMax H3 nodes: memory-efficient attention, scheduled tau with graph preview, and feed-forward chunking. Measured 1.14–1.44× vs SageAttention and −37% MLP peak VRAM on H3
—
Local home solar-plus-storage simulator: auto dispatch, 16-cell LFP pack, optional LLM hook
—
Generative models from scratch, progressively implemented in NumPy, C, and CUDA to uncover what ML frameworks hide.