OpenAI has published a blog post about how GPT-5.6 is being used to autonomously rewrite its own GPU kernels, the core code that executes the mathematical operations that make up the model. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Microsoft published the source code to its 6502 version of BASIC a while ago. It used to run on computers like the Apple II and the Commodore Pet. It has now been ported to C so that you can try it on a modern PC with Windows, Linux, or macOS. --- https://github.com/garyexplains/BASIC-M6502-CPORT/tree/main X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

TerraMaster TOS 7 has OpenClaw integrated directly into the OS including special skills for administering the NAS unit. It understands drives, networks shares, backups, system management, etc. In this video I show you how to set up and use OpenClaw on a TerraMaster F6-425 Pro. --- Amazon link for the F6-425 Pro: https://www.amazon.com/dp/B0H4FDN7ZW X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Discover the power of LLM Tool Calling (also known as function calling). In this video, you'll learn how Large Language Models can intelligently call external tools, APIs, and functions to solve real-world tasks, access live data, and use local computing power. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

OpenAI admits that its models caused "unprecedented cyber incident" by breaking out of its sandbox so that it could hack "Hugging Face’s production infrastructure to obtain test solutions directly from Hugging Face’s production database." --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

The Linux file system Btrfs includes several cool features including snapshots. You can use snapshots as part of your backup strategy and when combined with RAID then you are on the right path to fully protecting your data. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

SpaceXAI releases Grok Build as open source after major privacy blunder that ruined confidence in the service. Following yesterday's revelations that Grok Build was uploading code to a private Google Drive area, SpaceXAI has responded by making Grok Build open source. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Grok Build has been shown to package an entire repository (including files it was told not to open and secrets deleted from history) and upload it to xAI's private Google Cloud Storage. --- Links: https://cereblab.com/ and https://github.com/cereblab/grok-build-exfil-repro X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

How to train your own 10M parameter LLM | Get started with SerpApi using 250 free credits: https://serpapi.com/?utm_source=youtube&utm_campaign=garyexplains_july_2026 Here is everything you need to train your own 10 million parameter LLM using the complete works of Shakespeare. It works on a PC with or without a GPU. --- LLM Links: - https://github.com/angelos-p/llm-from-scratch/tree/main - https://github.com/garyexplains/examples/tree/master/tinynanogpt_chartok - https://github.com/garyexplains/examples/tree/master/tinystoriesgpt - https://huggingface.co/datasets/roneneldan/TinyStories/tree/main - https://huggingface.co/datasets/next-token/clean-PD-16000-books3 This video was sponsored by SerpApi. 00:00 Intro 02:07 10M parameter LLM trained on Shakespeare 07:40 24M parameter LLM trained on Tiny Stories 11:15 Wrap X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Squidbleed is a Heartbleed-style vulnerability that leaks internal memory from every version of Squid Proxy, in its default configuration. It was found by Claude Mythos Preview, and has now been fixed. --- Important links: https://blog.calif.io/p/squidbleed-cve-2026-47729 https://github.com/squid-cache/squid/security/advisories/GHSA-8c37-pxjq-qwrg X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Microsoft has just released the first version of WSL Containers. This new feature of Windows Subsystem for Linux enables you to run Linux containers directly on Windows without using Docker. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Tidal has introduced a new AI policy which demonetizes AI generated music. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Tidal has a new AI policy that demonetizes 100% AI-generated music. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Microsoft has built Rowhammer protection directly into the Azure Cobalt 200's memory controller. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Microsoft has built Rowhammer protection directly into the Azure Cobalt 200's memory controller. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Did you know Anthropic's Claude was named after the Bell Labs engineer, and MIT Professor, Claude Shannon? Shannon is often called the Father of Information Theory because of his work at the dawn of the digital computer age. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

I have got my very first number sequence published in the Online Encyclopedia of Integer Sequence (OEIS). A396572 is about a loop of different numbers where you keep doing XOR on neighboring pairs. The question is how many different arrangements you get before the pattern starts repeating. --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Watch me go from a simple business idea to a fully functional, professional consulting website (complete with contact forms, appointment booking, Stripe payments, and analytics) all using Rocket. --- Thanks to Rocket for sponsoring this video: https://www.rocket.new/?utm_source=youtube&utm_medium=garyexplains&utm_campaign=rocketbuild X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

In this video, I test DiffusionGemma, Google's new diffusion-based AI model. Instead of generating text one token at a time like ChatGPT, DiffusionGemma generates the entire response in parallel using diffusion, similar to image generation models. --- Thanks for MEGA for sponsoring this video: Get 50% off 3 TB of encrypted cloud storage https://mega.io/garyexplains?mct=garye --- X: https://twitter.com/garyexplains GitHub: https://github.com/garyexplains #garyexplains

Google has released a 12 billion parameter variant of its Gemma 4 series. The model only needs 16GB of RAM, or better still 16GB of VRAM, for a fast and capable model. --- ⭐ Please support my channel on Patreon! Get early access to videos, members-only content, behind-the-scenes updates, and join the Gary Explains Discord! Join here 👉 https://www.patreon.com/GaryExplains 🙌 Twitter: https://twitter.com/garyexplains Instagram: https://www.instagram.com/garyexplains/ #garyexplains

Microsoft has released the CoreUtils for Windows. The same commands you use on Linux or macOS, now work the same way on Windows! --- Thanks to Readdy AI for sponsoring this video: Link: https://bit.ly/Readdy4_Garyexplains Code: Gary #garyexplains

Arm has demonstrated a new mobile game that uses Arm Neural Technology and Unreal Engine MegaLights. The game is called Neural Dawn and it showcases the capabilities of the new tech coming in the next-generation of Arm Mali GPUs. --- ⭐ Please support my channel on Patreon! Get early access to videos, members-only content, behind-the-scenes updates, and join the Gary Explains Discord! Join here 👉 https://www.patreon.com/GaryExplains 🙌 Twitter: https://twitter.com/garyexplains Instagram: https://www.instagram.com/garyexplains/ #garyexplains

Microsoft has launched Intelligent Terminal, a fork of Windows Terminal with native agent integration. VEED: https://bit.ly/4tZDhVS --- Thanks to @VEEDSTUDIO for sponsoring this video Code: GARY30 (first 100 users get 30% off) 00:00 Intro 01:24 Getting started 03:01 Demo 05:20 Coding with Intelligent Terminal 07:26 Outro #garyexplains

The more data you give an AI model the better it can answer. In this video I show you how to grab the latest financial data about a company and then ask an LLM to analyze it for you. An easy way to grab the data is using Decodo Web Scraping API which does all the heavy lifting for you and can handle Javascript intensive pages, cookies, etc. -- Thanks to Decodo for sponsoring this video: Decodo: https://visit.decodo.com/yZ916B Decodo 10% discount code: GARYEXPLAINS #garyexplains

NVIDIA has announced a new range of PCs, laptops, and mini PCs using its own CPU and GPUs. The new NVIDIA DGX Station for Windows and the new NVIDIA RTX Spark want to reinvent Windows PCs for the age of personal AI! --- ⭐ Please support my channel on Patreon! Get early access to videos, members-only content, behind-the-scenes updates, and join the Gary Explains Discord! Join here 👉 https://www.patreon.com/GaryExplains 🙌 Twitter: https://twitter.com/garyexplains Instagram: https://www.instagram.com/garyexplains/ #garyexplains

In this video, I explain the key parameters that control how local large language models generate text, including temperature, top-k, top-p, min-p, repeat penalty, and other sampling options. --- Thanks to Genspark for sponsoring this video. 2026 Unlimited AI Chat & AI Image access, and Genspark is now offering a "GetStarted" bonus! New users can try premium features like AI web app building and deep research for free. Plus, you can earn extra credits just by completing simple tasks! https://www.genspark.ai/?utm_source=yt&utm_campaign=GaryExplains Genspark YouTube channel:@GensparkProduct #Genspark #WorkwithGenspark

A few years ago, turning an idea, a passion of yours, into a website was a complex thing to do. You needed backend code, frontend code, databases, authentication, deployment. But today, things are different. Today, we have Replit, where your creativity can run without limits. --- Get $10 worth of FREE credits: https://replit.com/refer/garysimsaa?trackingContext=universal-settings-modal #garyexplains

The problem for content creators today isn’t the lack of AI assisted tools, but that there are so many of them! Plus new ones are appearing all the time. It seems like, the moment you commit to one service a better tool appears somewhere else. Wery AI is an all-in-one platform that integrates dozens of AI image and video generators, like Runway, Pika, Veo, Sora, Nano Banana 2, Kling 3.0, and more. -- #ai #weryai #garyexplains

Bash hasn’t changed much in decades, so I built something different. Meet lunacmd: a Lua-powered command shell with pipes, scripting, and some unconventional ideas. --- ⭐ Please support my channel on Patreon! Get early access to videos, members-only content, behind-the-scenes updates, and join the Gary Explains Discord! Join here 👉 https://www.patreon.com/GaryExplains 🙌 Twitter: https://twitter.com/garyexplains Instagram: https://www.instagram.com/garyexplains/ #garyexplains

It is a simple question, can you replace Claude Code/Codex with OpenCode and a local LLM? 🤔 But it is quite hard to answer. In this video I test OpenCode with 3 different local AI models and see if they can do the same tasks as the big coding agents. --- ⭐ Please support my channel on Patreon! Get early access to videos, members-only content, behind-the-scenes updates, and join the Gary Explains Discord! Join here 👉 https://www.patreon.com/GaryExplains 🙌 Twitter: https://twitter.com/garyexplains Instagram: https://www.instagram.com/garyexplains/ #garyexplains

Is it possible to use tools like Codex or Clause Code with a local LLM. If you can the advantages would be massive: no expensive monthly subscription and no fears of running out of tokens. --- ⭐ Please support my channel on Patreon! Get early access to videos, members-only content, behind-the-scenes updates, and join the Gary Explains Discord! Join here 👉 https://www.patreon.com/GaryExplains 🙌 Twitter: https://twitter.com/garyexplains Instagram: https://www.instagram.com/garyexplains/ #garyexplains

Ubuntu 26.04, code named "Resolute Raccoon", was released a few days ago with GNOME 50, Linux kernel 7.0, and long term support. Here is a full desktop tour to help you get started. --- ⭐ Please support my channel on Patreon! Get early access to videos, members-only content, behind-the-scenes updates, and join the Gary Explains Discord! Join here 👉 https://www.patreon.com/GaryExplains 🙌 Twitter: https://twitter.com/garyexplains Instagram: https://www.instagram.com/garyexplains/ #garyexplains

Give Readdy.ai a try at https://bit.ly/Readdy4_Garyexplains and use my special code Gary for a discount. I gave an AI website builder just a business card… and asked it to create a full professional website from scratch. In this video, I tested Readdy.AI, an all-in-one AI-powered website builder that lets anyone create stunning company websites, landing pages, or promotional sites in minutes, even with zero coding skills. Watch as the AI turns a simple business card into a complete, modern website — Thanks to Readdy.Ai for sponsoring this video. 00:00 Intro 01:30 Demo using a plain language prompt 02:54 Create a website from a business card 03:30 Integrating Stripe payments 04:53 Publishing your website 07:00 SEO and driving traffic to your site 08:39 Creating a website from a template 11:00 Asking for help 11:30 Outro #garyexplains