56 subscribers
7.27K videos
7.88K links
Download Telegram
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ nvlabs/gateddeltanet-2

Gated DeltaNet-2: Smarter Linear Attention Explained

Linear attention models are great for efficiency, but they often struggle to edit their compressed memory without scrambling existing data. That is where Gated DeltaNet-2 changes the game. By decoupling the erase and write operations into independent, channel-wise gates, it allows the model to precisely control what information is removed from the key side and what new content is committed to the value side. This design leads to superior performance across language modeling and long-context retrieval tasks. Give this innovative architecture a look if you want to see how much better your models can remember.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ aakashadesara/ctop

Stop Managing AI Agents Manually: Meet CTOP

Are you juggling multiple AI coding agent sessions and losing track of what they are doing to your system resources? Meet the perfect solution for your terminal. This tool acts like a high-powered process viewer specifically designed for AI agents like Claude Code, Codex, and Devin. It offers a live, interactive dashboard where you can monitor real-time CPU usage, memory, API costs, and even context window tracking from a single pane. Whether you need to filter agent logs, sort by usage, or bulk-terminate stuck processes, this toolkit keeps your development environment organized and efficient. Take control of your AI workflows today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ darrenhinde/openagentscontrol

Take Control of AI Agents with OpenAgentsControl

Stop struggling with disorganized AI workflows and start building with precision. This repository introduces a powerful framework designed for plan-first development, ensuring your AI agents execute tasks only after you have reviewed and approved them. With built-in multi-language support for TypeScript, Python, Go, and Rust, it features an extensive system for automatic testing, code review, and validation. Whether you are managing complex subagents or streamlining your coding process, this tool provides the structure you need to maintain total control. Explore these sophisticated patterns to elevate your development efficiency and build smarter, reliable AI agents starting today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ maderix/ane

Training Neural Networks on Apple's Neural Engine

Have you ever wondered if you could train AI models directly on the Apple Neural Engine? Most people think this hardware is restricted to running inference, but this research project proves otherwise. By reverse-engineering private APIs, it bypasses standard limitations to perform forward and backward passes directly on Apple silicon. It achieves this by constructing custom compute graphs and using shared memory for zero-copy data pipelines. While it is not a production-ready framework, it serves as a fascinating proof of concept for edge AI optimization. If you are curious about pushing hardware boundaries, this project offers a unique look under the hood.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ openbmb/pilotdeck

Supercharge Your Productivity with PilotDeck AI Agents

Are you struggling to keep up with your daily digital tasks? Meet PilotDeck, a powerful, task-oriented platform designed to automate your workflow through AI agents. This system acts as a central hub, connecting various tools like note-taking apps, calendars, and even GitHub to handle complex requests automatically. It features a modular architecture that supports diverse integrations, allowing you to build custom skills that manage your projects and information efficiently. By streamlining how you interact with your favorite applications, PilotDeck saves you time and reduces manual effort. It is time to let your AI agents do the heavy lifting.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ tradesdontlie/tradingview-mcp

Automate Your TradingView Charts with AI

Imagine having an AI assistant that can instantly analyze your charts and manage your trading workspace for you. This repository provides a powerful bridge that connects your AI assistant to the TradingView Desktop application using standard debugging protocols. It enables you to automate your workflow by programmatically navigating timeframes, managing indicators, reading complex Pine Script data, and even conducting replay practice. By turning your trading interface into an interactive tool for your AI, this project helps you streamline analysis and focus on what matters most. Explore this bridge to level up your personal trading workflow today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ collabs-inc/collab-public

Build With AI Agents on an Infinite Canvas

Are you tired of constantly switching between tabs while working with AI agents? Meet Collaborator, an innovative desktop environment designed to streamline agentic development by bringing everything together in one place. It provides an infinite canvas where you can arrange your code editors, terminals, and markdown notes side by side, eliminating context switching and keeping your workflow fluid. Built with modern tools like React and Electron, it lets you visually organize your workspace, making it easier to manage complex projects with AI. It is a powerful, local-first solution for developers who want to build and iterate faster.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ revfactory/harness-100

Supercharge Your AI Workflows with This Massive Collection

Struggling to manage complex AI projects? This repository offers an extensive library of modular AI agent configurations designed to streamline everything from content creation and software development to data analysis and system architecture. Each folder provides a specialized toolkit, complete with predefined roles, expert agents, and technical skill sets tailored to specific tasks like building full-stack web apps, designing game narratives, or managing advertising campaigns. By leveraging these structured frameworks, you can quickly deploy high-level automation for your professional projects. Unlock a more efficient way to build and scale your digital assets with these ready-to-use AI workflows.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ alibaba/zvec

Build Lightning-Fast AI Apps with This Vector Database

Looking to boost your application performance? This repository offers a powerful, lightweight, and in-process vector database designed for lightning-fast operations. It tackles the challenges of high-speed data retrieval by providing robust implementations of advanced algorithms like HNSW, DiskANN, and IVF. By leveraging high-performance C plus plus code with SIMD optimizations, it allows developers to integrate efficient vector searching directly into their own systems without the overhead of external services. Whether you are building complex AI search features or scaling your data structures, this project delivers the core tools you need. Explore it today and take control of your data.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ ubikron/awesome-ai-osint

Master OSINT Investigations with AI

Are you struggling to keep up with the fast-evolving world of open-source intelligence? This repository is a goldmine for investigators looking to supercharge their workflows with artificial intelligence. It serves as an extensive, curated collection of articles, videos, and specialized tools dedicated to AI-driven OSINT. Whether you need help with image identification, geolocation, face analysis, or utilizing custom GPTs for complex research, this project provides the resources to get it done efficiently. Explore these resources to streamline your data gathering and analysis processes today, and stay ahead in the rapidly shifting landscape of modern digital investigations.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ aks-labs/circletosearch

Get Circle to Search on ANY Android Phone!

Want the convenience of screen search on your Android device without the restrictive ecosystem lock-in? This open-source project brings a powerful, privacy-focused alternative to your fingertips. It lets you circle or select any part of your screen to perform instant reverse image searches using multiple engines like Google, Bing, Yandex, and even ChatGPT. With offline OCR for text extraction and local QR scanning, it works entirely on-device to keep your data private. It is fully compatible with Android devices, offering a customizable, polished interface that restores control over your own search experience. Give it a try today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ jjang-ai/vmlx

Run Local AI Models on Apple Silicon with vMLX

Want to run high-performance AI models right on your Mac? Meet vMLX, a powerful self-hosted inference server built for Apple Silicon that lets you run LLMs, vision models, and image generators without needing third-party API keys. It offers an OpenAI and Anthropic compatible interface, making integration seamless. What makes it special is its sophisticated caching engine, featuring persistent disk storage, paged KV caching, and continuous batching for lightning-fast performance. Whether you use the command line or the native macOS desktop app, vMLX provides a robust, professional environment for local AI. Take control of your models and start building locally today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ crisng95/flowkit

Automate AI Video Creation with FlowKit

Stop creating AI videos manually and let FlowKit handle the heavy lifting for you. This standalone system acts as an intelligent bridge to the Google Flow API, automating everything from story conceptualization to final production. By using a clever Chrome extension as a secure gateway for authentication and reCAPTCHA, it orchestrates a complete pipeline that includes generating consistent reference images for characters, creating multi-scene video clips, adding narration, and even handling YouTube-optimized thumbnails. It solves the problem of visual inconsistency and manual tediousness, letting you produce cohesive, professional stories effortlessly. Take control of your workflow and start building better AI-generated content today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ ccplugins/awesome-claude-code-plugins

Supercharge Your Coding with Awesome Claude Code Plugins

Want to take your AI-powered development to the next level? This repository offers a massive collection of plugins designed to customize and extend your Claude Code experience. Whether you need specialized subagents for debugging, custom slash commands to streamline your workflow, or integrations for data and security tools, this curated list has you covered. By leveraging these lightweight extensions, you can automate repetitive tasks and keep your coding environment focused and efficient. Dive into these community-driven tools to simplify your architecture, documentation, and testing processes. Explore the library today to find the perfect productivity boost for your next project.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ nepfaff/scenesmith

Automated 3D Scene Generation with SceneSmith

Ever wanted to build a complex 3D environment just by describing it? Meet SceneSmith, a powerful tool designed for fully automated, agentic generation of simulation-ready indoor scenes. Simply provide a text prompt, and SceneSmith handles the rest, placing furniture and even inferring contextual details like placing ping pong paddles near a table. These generated scenes are fully separable with estimated physical properties, making them immediately ready for physics simulators. It uses advanced backends like SAM3D to create high-quality assets, giving researchers and developers a streamlined way to populate virtual worlds for robotics or simulation projects with zero human intervention.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ dair-ai/dair-academy-plugins

Supercharge Your Coding Assistant with DAIR Academy Plugins

Ever wish your AI coding assistant could do more than just write code? The DAIR Academy Plugins repository offers a powerful marketplace of open-source tools designed to extend your development environment. Whether you need to generate high-resolution images, orchestrate multiple AI models for better decision-making, or build automated research wikis, these plugins have you covered. You can even create structured learning paths or generate custom survey papers with ease. It is a fantastic way to customize your workflow and handle complex tasks directly inside your tools. Dive in today and start building smarter with these versatile, ready-to-use AI skills.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ amap-ml/dreamx-world

DreamX-World: Interactive AI Video Generation

Have you ever wanted to step inside your own AI-generated worlds? DreamX-World is a groundbreaking general-purpose model designed for interactive world simulation. By using advanced camera control and prompt-driven events, it lets you explore, control, and transform immersive environments in real-time. Whether it is realistic cityscapes or imaginative fantasy scenes, the model maintains high visual fidelity and consistent logic during long-horizon generation. It uses smart memory retrieval to keep scenes persistent as you revisit locations. Dive into a new era of creative exploration where you hold the power to shape your own evolving digital universe with ease.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ tele-ai/telestyle

Master Style Transfer with TeleStyle

Ever wanted to transform your videos or images into stunning new artistic styles while keeping the original content perfectly intact? Meet TeleStyle, a powerful model built on Qwen-Image-Edit designed to solve the tricky problem of feature entanglement in diffusion transformers. By using a unique curriculum learning framework, it masters diverse styles without losing content fidelity. It even includes a dedicated video-to-video module to ensure your stylized creations stay smooth and consistent. Whether you are a creator or a researcher, this lightweight tool offers state-of-the-art results for your visual projects. Give it a try to elevate your content today.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ u7079256/paperjury

Stop Submitting Before You Run This AI Stress-Test

Are you tired of getting rejected for avoidable mistakes? Stop guessing and start stress-testing your research with PaperJury. This tool transforms your paper into a closed-loop review process where AI acts as an adversarial jury. Instead of generic feedback, it uses a multi-stage engine to categorize issues into fixable, author-required, or invalid findings. It features deterministic guards for layout compliance and a courtroom-style trial to challenge your claims. You get a professional-grade audit that preserves your scientific judgment while catching overlooked flaws before submission. Take control of your research quality and ensure your paper is truly ready for prime time.

πŸ†” @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
πŸ“¦ sentient-agi/evoskill

Turn Your Coding Agent Into a Specialist with EvoSkill

Ever feel like your AI coding agent could be smarter? Meet EvoSkill, an open-source framework designed to automatically transform general AI agents into high-performing specialists. Instead of manual tweaking, this tool observes failed attempts and synthesizes reusable skills and optimized prompts to boost success rates. It works by running a feedback loop that evaluates new program variants against benchmarks, allowing you to iterate effectively. With support for major agents like Claude Code, OpenHands, and Goose, it is a powerful way to evolve your agent’s capabilities. Start evolving your coding workflow today and see what your AI can really achieve.

πŸ†” @hackernewsgithubprojects