Ever since I discovered the joy that is Rescueshark, it has become an essential part of my toolbox. My favorite use for it is putting operating systems in 'stasis,' which lets me swap between distros or go between Windows and Linux without losing any data.

After years of using ChatGPT and Claude, I'm finally starting to dabble in local LLMs. I'm not replacing cloud AI yet, but running Qwen2.5 or Llama 3.2 on my PC comes in handy when I don't want to hit message caps or censorship walls. You might think that running an LLM locally requires significant compute, but even an old GPU can handle several smaller models. I can't expect Claude-like intelligence from my local AI setup, but it's surprisingly good for general queries, document analysis, and productivity tasks, once I have the right tweaks dialed in. I was on the lookout for a cheap card to act as my dedicated local AI GPU, and was curious how low I could go. Fortunately, even an 8GB VRAM GPU that's several generations old has enough power to host 7B–8B models. I got a pre-owned RTX 3060 12GB, and it cost me much less than upgrading my primary GPU to one of the latest high-end models.

Connecting Claude to my apps sounded like a lot of work with not much payoff. I’d been using Spotify, Google Calendar, and Google Drive for years without much thought. They got the job done, and that’s what mattered to me. Then I finally connected them to Claude through its MCP-powered Connectors and realized I’d been using them in the most basic way possible. I was building playlists by hand, moving my calendar events one at a time, and going through my Google Drive files until I gave up. It was only after I connected them to Claude that I realized how much I’d been doing by hand when I didn’t need to.

Ever since Windows 11 was released, it has had a surprisingly large number of missing features. We're not talking about minor things, either; we're talking about things people got used to over decades of Windows use, only to vanish for seemingly no reason. To make matters worse, the AI boom caused Microsoft to invest a ton of money and time into Copilot, as the core Windows 11 experience was pushed aside.

It finally happens: a smart plug goes offline, and you're not sure which one. You open your router dashboard and you're greeted by a list of devices with names like "ESP_88bc," "Android_4f91," and three separate entries that just say "Amazon Echo." You know one of those plugs controls the lamp in the office, but which IP address belongs to which physical device is anybody's guess. What should take two minutes turns into twenty, and you haven't even touched a cable yet. Your gear can be the smartest, your setup can be the sickest, but it's only as organized as you make it. Taking the time to label things both physically and digitally saves so much grief later on, and it should be the first thing you do when you first set up your smart home.

Valve is known for a lot of masterstrokes in the PC gaming space. With its unrivaled storefront, iconic game franchises, SteamOS, the Steam Deck, and even the company culture, Valve is a bit of a unicorn. The upcoming Steam Machine is set to join the Steam Deck and Steam Controller in Valve's gaming hardware lineup. While Valve's past performance, both in hardware and software, has been impeccable, there's also the original "Steam Machines" blot on its report card. The new Steam Machine is a curious device, marketed not as a console but as a PC running SteamOS and powering 4K 60 FPS gaming. Its underpowered internals were already cause for alarm, but even if you assume Valve will nail the optimization, the target consumer for the Steam Machine is a bit up in the air. We know that Valve has confirmed a 2026 launch, but with worse pricing than that of consoles and hardware that can't rival budget PCs, who is the Steam Machine really for?

Some home lab services earn a dangerous reputation for being simple. You install them, point them at the right devices, confirm the dashboard loads, and then mentally file them under “handled.” That sounds comforting, especially when your home lab already has enough blinking lights, dashboards, containers, and mystery alerts demanding attention. The problem is that the services most often described as “set and forget” are usually the ones quietly sitting closest to the foundation.

This is The Stepback, a weekly newsletter breaking down one essential story from the tech world. For more news about how AI is seeping into our daily lives, follow Emma Roth. The Stepback arrives in our subscribers' inboxes at 8AM ET. Opt in for The Stepback here. How it startedIn 2021, McDonald's became one of the first major fast-food chains to greet customers with an AI chatbot at the drive-thru. It started small, deploying the voice-ordering technology at 10 of its locations in Chicago. McDonald's developed its drive-thru tech after acquiring Apprente, a startup focused on voice-based, conversational technology in 2019, and later wor …Read the full story at The Verge.

Every tech enthusiast in the world has grown up discussing the best GPUs of their time. For some, it was the GeForce 3 at the beginning of this century. The Radeon HD 5870 is spoken of with reverence, too. There's the GTX 1080 Ti, of course, but one card that's often left out of the conversation is an Nvidia GPU that truly deserves a place in the annals of gaming and tech history — the GTX Titan.

When you're busy at work and need a document ASAP, the last thing you want is a buggy, slow preview pane when you receive a file. Unfortunately, if you use Microsoft Teams to share and receive Office files, you've likely experienced this first-hand with previews that take too long to load, hog resources, or sometimes even fail to load at all. Well, here's some good news: not only does Microsoft know the problem exists, but it'll publish an update soon that should fix things.

Package managers are one of the best parts of Linux, making it easier to manage what's installed on your system, including specific versions of software you may need for specific projects. But they come with the downside that different distros often use different package managers, so installing the tools you need on a new PC with a new distro, or making the jump to a different flavor of Linux, can come with more hurdles than you'd like.

I recently started integrating local LLMs with my arsenal of free and open-source tools, and they’ve been a game-changer for my productivity needs. Whether it’s generating precise OCR scans or helping me rewrite long snippets of code in the right indentation, self-hosted models are surprisingly capable at automating everyday tasks. What’s more, the FOSS ecosystem has tons of obscure AI tools that are productivity powerhouses – provided you use them for the right tasks.