AMD just made a move that has the FPGA community up in arms. The company’s Vivado 2026.1 release drops Linux support from the free tier of its FPGA development tools. If you’re a hobbyist, student, or small shop working on Xilinx FPGAs on Linux, you now need to pay for a license or stick with an older version.

The AMD support forum thread shows the kind of reaction you’d expect. This isn’t some obscure edge case. FPGAs are a hardware developer’s bread and butter, and Vivado is how you program AMD’s (formerly Xilinx) chips. A lot of that work happens on Linux.

The practical impact: if you’re doing FPGA work professionally, you probably already have a license. But if you’re learning, teaching, or tinkering, you’re stuck. Either pay up, switch to Windows, or stay on 2025 releases until something breaks.

AMD hasn’t offered much explanation. The forum post is light on details, and the company’s messaging so far hasn’t clarified whether this is a cost-cutting move, a strategic shift, or just bad planning. What’s clear is that it’s alienating exactly the kind of users who build expertise on your platform and eventually become paying customers.

Meanwhile, Microsoft digs through the archives

In better news for people who care about computing history, Microsoft just open-sourced what they’re calling the earliest DOS source code ever found. This isn’t MS-DOS 1.0. It’s an internal development version that predates the first commercial release.

The code went up on GitHub after being discovered in Microsoft’s own archives. If you’ve ever wondered what the very first versions of DOS looked like before they shipped, here’s your chance to dig in.

This won’t matter to most working developers, but it’s a genuine artifact. Early DOS shaped how millions of people used computers, and having the source available means researchers and historians can actually study it instead of guessing.

Microsoft has been slowly open-sourcing old software over the past few years. It’s a good trend. These aren’t competitive products anymore. They’re historical records, and making them public helps everyone understand how we got here.

The tools roundup

A few other things crossed the radar this week:

Simon Willison wrote about the <dl> element, pulling from Ben Meyer’s article on how description lists actually work in HTML. Turns out you can nest multiple <dd> elements under one <dt>, wrap them in a <div> for styling, and use ARIA labels for accessibility. Also, they’ve been called “description lists” since 2008, not “definition lists.” If you write HTML for a living and didn’t know that, you’re not alone.

Ken Shirriff reverse-engineered a Spacelab computer from 1980. The full writeup walks through the hardware architecture, the custom chips, and how NASA designed computers for spaceflight four decades ago. If you like hardware teardowns and space history, it’s worth your time.

Andrew Ayer disassembled the 80386 microcode. The blog post goes deep on how Intel implemented instructions at the microcode level in one of the most influential x86 processors ever made. This is extremely niche, but if you care about low-level CPU internals, it’s fascinating.

What matters here

The Vivado story matters because it’s a reminder that “free tier” isn’t a guarantee. Companies change policies, and if your workflow depends on vendor tooling, you’re at their mercy. That’s not new, but it’s always worth remembering.

The DOS release matters because it’s a piece of history that would’ve stayed locked up otherwise. Open-sourcing old software doesn’t cost much, and it helps people learn.

The rest is solid technical writing from people who care about their craft. Not everything has to be breaking news. Sometimes it’s just good to read about how things work.