Hasty Briefsbeta

Bilingual

GitHub - mattpocock/skills: My personal directory of skills, straight from my .claude directory.
a month ago
- Agent skills extend capabilities in planning, development, and tooling, aiding problem-solving before coding.
- Planning skills include to-prd, to-issues, grill-me, design-an-interface, and request-refactor-plan for tasks like creating PRDs and interviewing.
- Development skills cover tdd, triage-issue, improve-codebase-architecture, migrate-to-shoehorn, scaffold-exercises, setup-pre-commit, git-guardrails-claude-code, write-a-skill, edit-article, ubiquitous-language, and obsidian-vault.
GitHub - microsoft/typescript-go: Staging repo for development of native port of TypeScript
a month ago
- Announcement of TypeScript native preview build on npm as @typescript/native-preview, installable via npm install.
- Preview VS Code extension available; enable by setting 'js/ts.experimental.useTsgo': true in VS Code settings.
- Feature status table provided for TypeScript Native, including program creation, parsing, type checking as done, while JavaScript-specific inference, declaration emit, and language service are in progress.
- Project is a work in progress, not at full feature parity; users advised to check list before logging issues.
- Long-term plan to merge into microsoft/TypeScript; repo and issue tracker will eventually be closed.
- Project welcomes contributions, requires CLA, and follows Microsoft Open Source Code of Conduct and trademark guidelines.
GitHub - deepseek-ai/DeepEP: DeepEP: an efficient expert-parallel communication library
a month ago
- DeepEP is a communication library for Mixture-of-Experts (MoE) and expert parallelism, offering high-throughput, low-latency all-to-all GPU kernels (dispatch and combine) with low-precision support like FP8.
- The library includes kernels optimized for asymmetric-domain bandwidth forwarding for group-limited gating (e.g., NVLink to RDMA) and low-latency kernels with pure RDMA for inference decoding, along with an SM-free communication-computation overlapping method.
- Performance tests on H800 GPUs show intranode and internode bandwidths up to ~158 GB/s (NVLink) and ~58 GB/s (RDMA) for normal kernels, and latencies as low as 77 us for low-latency kernels.
- Recent optimizations (2025) from Tencent and others improved performance by up to 30%, enhanced NVLink usage, and added features like multi-QP support and AMD GPU compatibility via backends like MORI.
- System requirements include Ampere/Hopper GPUs, Python 3.8+, specific CUDA versions, PyTorch 2.1+, NVLink, RDMA network, and NVSHMEM dependency.
- Usage examples cover normal kernels for training/inference prefilling (with SM control and buffer management) and low-latency kernels for inference decoding (with hook-based overlapping).
- Advanced features include traffic isolation via InfiniBand Virtual Lanes, adaptive routing recommendations, and optimizations like undefined-behavior PTX instructions for performance (can be disabled).
- The library is MIT-licensed (except NVSHMEM references) and encourages citation; contributions include zero-copy, eager protocols, hybrid-EP, and diagnostic tools.
GitHub - codecrafters-io/build-your-own-x: Master programming by recreating your favorite technologies from scratch.
a month ago
- Repository with step-by-step guides for recreating technologies from scratch to aid learning.
- Includes a wide range of projects such as 3D Renderers, AI Models, Blockchain, Databases, Games, Operating Systems, and Web Servers.
- Features guides in multiple programming languages including C, C++, Python, JavaScript, Go, Rust, and more.
- Contains both software and hardware projects, like building a VR headset or a processor from scratch.
- Encourages community contributions through PRs and issue creation, with open copyright under CodeCrafters, Inc.
GitHub - dani-garcia/vaultwarden: Unofficial Bitwarden compatible server written in Rust, formerly known as bitwarden_rs
a month ago
- Vaultwarden is an alternative server implementation of the Bitwarden Client API, written in Rust, for self-hosted deployment.
- It offers a nearly complete Bitwarden Client API, including features like personal vaults, organizations, two-factor authentication, and a web vault.
- Installation is recommended via container images (e.g., Docker) with persistent storage and a reverse proxy for security.
- Users should report bugs directly to the Vaultwarden community, not to official Bitwarden support.
- The project was renamed from Bitwarden_RS to Vaultwarden to avoid confusion and trademark issues.
- Maintainers focus on self-hosting for individuals and small organizations, with no liability for data loss.
GitHub - google/osv-scanner: Vulnerability scanner written in Go which uses the data provided by https://osv.dev
a month ago
- OSV-Scanner is a CLI tool that scans project dependencies to find known vulnerabilities using the OSV database.
- It supports a wide range of languages (e.g., C/C++, Go, Java, Python, Rust), package managers (e.g., npm, pip, maven), operating systems, and container images.
- The tool offers features like call analysis to reduce false positives, guided remediation for version upgrades, and offline scanning with a local database.
- OSV-Scanner queries external services such as api.osv.dev for vulnerabilities and deps.dev for package information, with options for offline use.
- Guided remediation is experimental, supports npm and Maven ecosystems, and carries risks like executing scripts from untrusted projects.
- Users can install via prebuilt binaries or build from source, with detailed documentation available for usage and contribution.
GitHub - chiphuyen/aie-book: [WIP] Resources for AI engineers. Also contains supporting materials for the book AI Engineering (Chip Huyen, 2025)
a month ago
- The availability of foundation models has made AI accessible as a development tool for all.
- The book provides an end-to-end framework for adapting foundation models, including LLMs and LMMs, to real-world problems.
- It addresses key questions in AI application development, such as evaluation, hallucinations, prompt engineering, RAG, agents, finetuning, data validation, efficiency, and security.
- The content is based on case studies, references, and expert reviews, drawing from the author's decade of experience in ML and language models.
- This book focuses on fundamentals of AI engineering rather than specific tools, ensuring lasting relevance.
- It complements 'Designing Machine Learning Systems' (DMLS), covering foundation models while DMLS focuses on traditional ML, with both being self-contained.
- Target audience includes AI engineers, data scientists, managers, and product managers involved in building or optimizing AI applications.
- The book also benefits tool developers, researchers, job candidates, and those seeking to understand AI's capabilities and limitations.
- Technical depth is provided in some sections, with warnings for readers to skip if desired.
- It includes praise from industry experts and acknowledges numerous contributors for their feedback and support.
GitHub - coreyhaines31/marketingskills: Marketing skills for Claude Code and AI agents. CRO, copywriting, SEO, analytics, and growth engineering.
a month ago
- A collection of AI agent skills for marketing tasks designed for technical marketers and founders.
- Skills provide specialized knowledge and workflows for conversion optimization, copywriting, SEO, analytics, and growth engineering.
- Works with various AI coding agents like Claude Code, OpenAI Codex, Cursor, Windsurf, and others supporting the Agent Skills spec.
- Includes foundational skill 'product-marketing-context' that other skills reference to understand product, audience, and positioning.
- Skills are cross-referenced and categorized into SEO & Content, CRO, Content & Copy, Paid & Measurement, Growth & Retention, Sales & GTM, and Strategy.
- Installation options: using npx skills, Claude Code's plugin system, cloning the repo, or using SkillKit for multi-agent support.
- Skills can be invoked directly or used when agents recognize marketing tasks, such as optimizing landing pages or writing copy.
GitHub - mksglu/context-mode: Context window optimization for AI coding agents. Sandboxes tool output, 98% reduction. 12 platforms
a month ago
- Context Mode is an MCP server that addresses context window issues by keeping raw data out of the context window using sandbox tools, achieving up to 98% reduction in context usage.
- It ensures session continuity by tracking all session events (file edits, git operations, tasks, errors, user decisions) in SQLite and using BM25 search via FTS5 to restore state after conversation compaction, allowing the model to resume where it left off.
- The tool promotes a 'Think in Code' paradigm, encouraging LLMs to generate code for analysis instead of processing data directly, which reduces tool calls and saves context.
- Installation and configuration vary across 12 platforms (e.g., Claude Code, Gemini CLI, VS Code Copilot, Cursor), with differences in hook support, routing enforcement, and session continuity capabilities.
- Key tools include ctx_execute, ctx_batch_execute, ctx_index, ctx_search, and ctx_fetch_and_index, which process data in isolation and return only relevant outputs.
- Session continuity relies on hooks like PreToolUse, PostToolUse, PreCompact, and SessionStart to capture events, build snapshots, and restore state, with varying support across platforms.
- Context Mode operates locally with no telemetry, uses permission rules for security, and is licensed under Elastic License 2.0, ensuring privacy and source availability.
GitHub - microsoft/onnxruntime: ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
a month ago
- ONNX Runtime is a cross-platform accelerator for machine learning inference and training.
- It supports models from various frameworks like PyTorch, TensorFlow/Keras, and libraries such as scikit-learn, LightGBM, and XGBoost.
- It offers faster performance and lower costs by leveraging hardware accelerators, graph optimizations, and transforms, and is compatible with multiple hardware and operating systems.
- Training acceleration is available for transformer models on multi-node NVIDIA GPUs with minimal code changes in existing PyTorch scripts.
- Resources include documentation, tutorials, sample repositories, release details on GitHub, and a roadmap for upcoming features.
- The project collects usage data on Windows for product improvement, welcomes contributions, and is licensed under the MIT License with a code of conduct.
GitHub - Alishahryar1/free-claude-code: Use claude-code for free in the terminal, VSCode extension or via discord like openclaw
a month ago
- Free Claude Code is a lightweight proxy that routes Claude Code's Anthropic API calls to multiple providers, including NVIDIA NIM (free 40 requests/minute), OpenRouter, DeepSeek, LM Studio, and llama.cpp, acting as a drop-in replacement with minimal setup.
- It supports per-model mapping, allowing specific Claude models (Opus, Sonnet, Haiku) to be routed to different providers and models, and includes features like thinking token support, heuristic tool parsing, request optimization to intercept trivial API calls, and smart rate limiting.
- Setup involves obtaining API keys for desired providers, configuring environment variables in a .env file, and starting the proxy server; the Claude Code CLI or VSCode extension can then be pointed to the proxy with two environment variables.
- Additional features include a Discord/Telegram bot for remote autonomous coding with session persistence and voice note support, extensible architecture for adding new providers or messaging platforms, and optional authentication to restrict proxy access.
- The proxy translates requests from Anthropic format to OpenAI-compatible formats, optimizes performance by locally handling trivial requests, and supports various local and cloud-based LLM providers, with configuration options for rate limiting, concurrency, and voice transcription.
GitHub - Anil-matcha/Open-Generative-AI: Uncensored, open-source alternative to Higgsfield AI, Freepik, Krea, Openart AI — Free, unrestricted AI image & video generation studio with 200+ models (Flux,
a month ago
- Open Generative AI is a free, open-source alternative to Higgsfield AI, Freepik, Krea, and Openart AI, with no content filters, subscription fees, or closed ecosystem.
- It offers 200+ AI models for image and video generation, including text-to-image, image-to-image, text-to-video, image-to-video, and lip sync capabilities.
- The platform supports multi-image input (up to 14 reference images), local inference via stable-diffusion.cpp, and self-hosting with full data privacy.
- Key features include Image Studio, Video Studio, Lip Sync Studio, Cinema Studio with camera controls, and Workflow Studio for building multi-step AI pipelines.
- It provides desktop apps for macOS, Windows, and Linux, with a hosted web version available at muapi.ai/open-generative-ai, and is MIT-licensed for customization.
GitHub - huggingface/ml-intern: 🤗 ml-intern: an open-source ML engineer that reads papers, trains models, and ships ML models
a month ago
- An ML intern tool autonomously researches, writes, and ships ML code using Hugging Face ecosystem.
- Setup involves cloning a repository, installing dependencies, and configuring API keys in a .env file.
- Supports interactive and headless modes with options for model selection, iteration limits, and streaming control.
- Architecture includes a submission loop, agentic loop with context management, tool routing, and doom loop detection.
- Event system emits various states like processing, tool calls, outputs, errors, and approvals during execution.
- Tools can be customized by editing agent/core/tools.py, and configuration via configs/main_agent_config.json.
GitHub - vercel-labs/skills: The open agent skills tool - npx skills
a month ago
- The CLI tool manages agent skills, which are reusable instruction sets defined in SKILL.md files to extend coding agent capabilities.
- Supports adding skills from various sources: GitHub shorthand, full URLs, GitLab, git URLs, and local paths.
- Provides multiple commands: add, list, find, remove, update, and init for skill management.
- Installation scopes include project-specific (default) or global using -g flag, with symlink or copy methods.
- Lists numerous compatible agents like Claude Code, OpenCode, Cursor, Codex, and many others with specific installation paths.
- Skills require a YAML frontmatter with name and description; can be marked as internal with metadata.internal: true.
- CLI searches for skills in standard directories within repositories and supports Claude Code plugin marketplace compatibility.
- Includes troubleshooting tips for installation issues, environment variables for internal skills, and telemetry details.
- Offers comprehensive documentation links for each supported agent and the Agent Skills Specification.
GitHub - Z4nzu/hackingtool: ALL IN ONE Hacking Tool For Hackers
a month ago
- HackingTool is a command-line interface (CLI) tool offering various cybersecurity commands and tool categories for security testing.
- The tool provides commands like /query for searching, t for tags, r for recommendations, and ? for help, with options to install, quit, or navigate back.
- It includes 20 categories such as Information Gathering, Web Attack, Phishing, Forensics, and Active Directory, each listing specific tools like nmap, Burp Suite, and Wireshark.
- Installation methods include a shell script, manual cloning with Python, Docker builds, and Docker Compose, with dependencies on Python 3.10+, Go 1.21+, Ruby, and Docker.
- The tool emphasizes ethical use for authorized security testing and encourages community contributions through issue templates and pull requests.
GitHub - AIDC-AI/Pixelle-Video: 🚀 AI 全自动短视频引擎 | AI Fully Automated Short Video Engine
a month ago
- Pixelle-Video is an automated video generation tool that requires only a topic to produce a complete video.
- It automates scriptwriting, AI-generated images/videos, voice narration, background music addition, and video synthesis.
- The tool features modular design with a clear workflow: script generation, image planning, audio processing, and video synthesis.
- Recent updates include motion migration, digital human narration, image-to-video pipelines, multi-language TTS support, and enhanced API configurations.
- It supports various AI models, flexible customization, multiple templates, and aspect ratios, and can run on local or cloud setups.
- Setup is simple with a Windows integrated package or manual installation via Python, uv, and ffmpeg.
- The web interface includes system configuration, content generation options, audio settings, image and video customization, and real-time progress tracking.
- Common questions address generation time, improving video quality, and cost-effective usage scenarios.
- It is inspired by open-source projects like Pixelle-MCP, MoneyPrinterTurbo, and ComfyKit, and is licensed under Apache 2.0.
GitHub - open-metadata/OpenMetadata: OpenMetadata is a unified metadata platform for data discovery, data observability, and data governance powered by a central metadata repository, in-depth column l
a month ago
- OpenMetadata is a unified metadata platform for data discovery, observability, and governance, powered by a central repository.
- Its four main components include Metadata Schemas, Metadata Store, Metadata APIs, and an Ingestion Framework with over 84 connectors.
- Key features include data discovery, collaboration, quality monitoring, governance, insights, lineage, documentation, observability, security, and webhooks integration.
- It is an open-source project with a vibrant community, Apache 2.0 license, and supports sandbox testing and contributions.
GitHub - langfuse/langfuse: 🪢 Open source LLM engineering platform: LLM Observability, metrics, evals, prompt management, playground, datasets. Integrates with OpenTelemetry, Langchain, OpenAI SDK, Li
a month ago
- Langfuse is an open-source LLM engineering platform for collaborative development, monitoring, and debugging of AI applications, with quick self-hosting options.
- Key features include LLM application observability for tracking calls and debugging logs, prompt management with version control, flexible evaluations, datasets for testing, and an LLM playground for iteration.
- It offers comprehensive APIs and SDKs, supports deployment via Docker, VMs, or Kubernetes, and integrates with various frameworks like LangChain, OpenAI, and LlamaIndex.
- The platform includes a quickstart guide for instrumentation using Python decorators, and provides support through documentation, GitHub Discussions, and an in-app chat widget.
GitHub - HKUDS/RAG-Anything: "RAG-Anything: All-in-One RAG Framework"
a month ago
- RAG-Anything is an All-in-One Multimodal Document Processing RAG system built on LightRAG, designed to handle diverse content like text, images, tables, and equations in modern documents.
- Key features include an End-to-End Multimodal Pipeline, Universal Document Support, Specialized Content Analysis, a Multimodal Knowledge Graph, Adaptive Processing Modes, Direct Content List Insertion, and Hybrid Intelligent Retrieval.
- The system utilizes a multi-stage pipeline with components like MinerU Integration, Adaptive Content Decomposition, and Universal Format Support, along with modality-aware processors for visual content, structured data, and mathematical expressions.
- It offers multiple query methods: Pure Text Queries, VLM Enhanced Queries for automatic image analysis, and Multimodal Queries for enhanced queries with specific content analysis.
- Installation is possible via pip (e.g., 'pip install raganything') with optional dependencies for extended format support, and it supports integration with existing LightRAG instances and direct insertion of pre-parsed content lists.
- RAG-Anything supports multiple parsers (MinerU, Docling, PaddleOCR), various document and content types, and includes practical demos in the examples/ directory, with a technical report released and community milestones like reaching 1k+ GitHub stars.
GitHub - dayanch96/YTLite: A flexible enhancer for YouTube on iOS
a month ago
- YouTube Plus is a flexible enhancer for iOS with over 100 customizable options, but version 5.2 onwards requires a subscription, with the last free version being 5.2b4.
- To build the YouTube Plus app via GitHub Actions, users must fork the repo, enable permissions, sync updates, and run a workflow with steps like uploading a decrypted IPA file and selecting tweaks.
- The build process supports options for using the latest release or a beta tweak file, with detailed notes on providing direct download links and downloading the final app from the forked repo's releases.
- This repository includes integration details for tweaks like YouPiP (enables Picture-in-Picture), YTUHD (unlocks 2K/4K resolutions), Return YouTube Dislikes (restores dislike counts), YouQuality (video quality controls), and DontEatMyContent (prevents notch interference).
- Each tweak has preferences accessible in YouTube settings and open-source code available on the developers' GitHub repositories, with the latest supported YouTube version confirmed as 21.15.5 tested on April 20, 2026.

first prev1next

About|Login

Hasty Briefsbeta

GitHub - mattpocock/skills: My personal directory of skills, straight from my .claude directory.

GitHub - microsoft/typescript-go: Staging repo for development of native port of TypeScript

GitHub - deepseek-ai/DeepEP: DeepEP: an efficient expert-parallel communication library

GitHub - codecrafters-io/build-your-own-x: Master programming by recreating your favorite technologies from scratch.

GitHub - dani-garcia/vaultwarden: Unofficial Bitwarden compatible server written in Rust, formerly known as bitwarden_rs

GitHub - google/osv-scanner: Vulnerability scanner written in Go which uses the data provided by https://osv.dev

GitHub - chiphuyen/aie-book: [WIP] Resources for AI engineers. Also contains supporting materials for the book AI Engineering (Chip Huyen, 2025)

GitHub - coreyhaines31/marketingskills: Marketing skills for Claude Code and AI agents. CRO, copywriting, SEO, analytics, and growth engineering.

GitHub - mksglu/context-mode: Context window optimization for AI coding agents. Sandboxes tool output, 98% reduction. 12 platforms

GitHub - microsoft/onnxruntime: ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator

GitHub - Alishahryar1/free-claude-code: Use claude-code for free in the terminal, VSCode extension or via discord like openclaw

GitHub - Anil-matcha/Open-Generative-AI: Uncensored, open-source alternative to Higgsfield AI, Freepik, Krea, Openart AI — Free, unrestricted AI image & video generation studio with 200+ models (Flux,

GitHub - huggingface/ml-intern: 🤗 ml-intern: an open-source ML engineer that reads papers, trains models, and ships ML models

GitHub - vercel-labs/skills: The open agent skills tool - npx skills

GitHub - Z4nzu/hackingtool: ALL IN ONE Hacking Tool For Hackers

GitHub - AIDC-AI/Pixelle-Video: 🚀 AI 全自动短视频引擎 | AI Fully Automated Short Video Engine

GitHub - open-metadata/OpenMetadata: OpenMetadata is a unified metadata platform for data discovery, data observability, and data governance powered by a central metadata repository, in-depth column l

GitHub - langfuse/langfuse: 🪢 Open source LLM engineering platform: LLM Observability, metrics, evals, prompt management, playground, datasets. Integrates with OpenTelemetry, Langchain, OpenAI SDK, Li

GitHub - HKUDS/RAG-Anything: "RAG-Anything: All-in-One RAG Framework"

GitHub - dayanch96/YTLite: A flexible enhancer for YouTube on iOS