Speculative Decoding in vLLM on AMD GPUs
A practical guide to speculative decoding in vLLM on AMD GPUs, covering draft-and-verify mechanics, MTP, EAGLE-3, DFlash, DSpark, configuration, tuning, and ben
The latest blog posts to take off on Hacker News. 80,792 posts.
A practical guide to speculative decoding in vLLM on AMD GPUs, covering draft-and-verify mechanics, MTP, EAGLE-3, DFlash, DSpark, configuration, tuning, and ben
Computer Science, Machine Learning, Programming, Art, Mathematics, Philosophy, and Short Fiction
When I started programming, my whole day revolved around development. I’d wake up, write code, and go to sleep. And that’s how it was day in and day out. My entire social circle consisted of people who also wrote code. I studied frameworks, libraries, and various programming languages.
“Universal access to all human knowledge is within our grasp. Our job is to put the best our world has to offer within the reach of our children.” —Brewster Kahle, Internet Archive Digital Librarian Every time you search the Wayback Machine or explore a collection at the Internet Archive, you are accessing a global…
I challenged myself to make a tiny subset of Python in C.
Math as cultural institution.
After a year of work, hundreds of commits, 5 generations of pull requests, dozens of tests, rewriting part of Anubis in Rust, the first compiler bug of my career, and at least three times making my tower run out of ram I think I have finally done it. The next version of Anubis will ship with WebAssembly-based proof of…
A couple years ago, GM made waves by saying that soon, all their new cars would be completely free of Apple's CarPlay and Android Auto support, instead opting for the "Android Automotive OS" in their cars. Admittedly, I'm a huge fan of CarPlay for basic safety reasons and
Jakub Pachocki reflects on increasingly capable AI and the challenge of keeping it aligned. He calls for stronger safeguards and international coordination.
The NetBSD Project is pleased to announce NetBSD 9.5, the fifth (and final) release from the NetBSD 9 stable branch.
Inside OpenAI, coding agents are reshaping AI research. Explore early data on agent usage, experiment velocity, task complexity, and research acceleration.
It's complicated
The day we discovered we had been designated a global terrorist organization, we promised that we would try to resist as long as this was possible, that we would not back down as long as we could see options to stand our ground. For years, we have proudly maintained and defended a free, privacy-friendly, autonomous…
Porting Linux to Apple Silicon
The best ice fishing in the Solar System is worth finding
I need to make a painful confession: somehow, LinkedIn has become an important social network to me. This isn’t necessarily due to LinkedIn’s sparkling competence, of course. To the contrary, LinkedIn is the Gerald Ford of social networks: the normal one that was left standing as the Richard Nixons and the Spiro…
James Marriott’s The New Dark Ages
In the last 28 years of publishing software, I’ve had three distinct eras of software licensing. All of my recent stuff is available under the European Union Public License 1.2, and I thought to explain why.
Modern programming languages are full of violence, vulgar abbreviations, and naked imperatives. Bespoke is a statically typed alternative in which exceptions are regrettable circumstances, mutexes ...
By enabling access to space, we contribute to humanity’s progress and our planet’s sustainable technological and economic development. Select your journey to orbit and launch with us.
My previous AMD Based FreeBSD Desktop is still up and running – it will now be my kids computer … so I needed a new one … but not a bigger one :) I spent way too much time finding…
It’s very tempting to imagine that AI turns everyone into a tool-builder - now everyone can just ask the model to make the software they need, and apps as we know them are dead. I think that misunderstands how most people think and where software actually comes from, and more importantly, it isn’t
Your open-source personal cloud: containerized apps, unified auth, good UX. Self-hosting should feel like using a smartphone that serves webapps, not a sysadmin side job.
Six years ago I published a blog post Chrome exempts Google sites from user site data settings that received quite a bit of attention: Michael Tsai , Hacker News The Register , The Verge , and Gizmodo , among others. The blog post was about a Google Chrome bug that mysteriously exempted Google-owned sites from the…
Reading is important to me. While I’m not a quick reader nor an especially voracious one, I have found that long-form reading has had a profound influence on me over my life.
Labor Day weekend has always been the back to school marker for me. I'm 60, closing in on 61, and thinking a lot more about what I actually care about. Amy texted me a list yesterday.
Public Health professor discusses how ultra-processed foods are a direct legacy of the tobacco industry, and are responsible for a dramatic rise in obesity, diabetes and other chronic diseases across the country.
U.S. workers form first recognized unit as global union movement builds across Wikimedia Foundation
Metaphors like "a sinking ship" are often used to describe codebases, but are misleading. A _business_ will sink long before code quality reaches a hypothetical floor. Technical debt has no bankruptcy, no clean reset, so metaphors that imply an end provide a false sense of security.
Ok it’s not exactly $60 but you can get the board itself for that amount of money on eBay if you’re lucky, but on average the price is around $70-$100 (as of November 2025) which is still very cheap considering you can play Cyberpunk 2077 on this.