Artem Golubin

Why am I getting spammed from IBM e-mails about IBM repos? Have you lost your mind @IBMDeveloper @IBM? You scraped that e-mail from GitHub. I only get spam there.

@strager @InSession314 It's very short, give it a read tonsky.me/blog/fall-of-n…

@rushter @InSession314 Yes, I didn't read the article 3 quote tweets down. Is it arguing in favor of improving native development (craft), or is it giving up on improving things and just saying 'Electron is good enough so let's use it'?

Does nobody care about their craft anymore???

@InSession314 @strager You did not even read the article that they discuss. Same for the @strager I guess. It has nothing to do with craft.

These guys aren’t programmers. They’re focused on churning out SAAS slop so they can make money. For them it isn’t their “craft” it’s more akin to their job, and yeah like most people they probably don’t give a shit about their job. Their opinions should be discarded and we need to split off the YCombinator modern Silicon Valley VC machine in the public consciousness from people who actually give a shit lol. Lots of cool projects out there, but outside of the occasional “Google upset with FFMPEG” drama you don’t really hear from them in gen pop or on the algorithms cause they’re more niche. Everyone’s got a hot take on the latest nonsense OpenAI or Anthropic puts out, it’s hard to get a bunch of people to give a shit about creating the perfect GFX plugin for N64 emulation. Easy to get black pilled when you scrolls through the twitter timelines but these people are untimely not serious.

I reported a malicious repository to GitHub. Three days later, still no action from GitHub 😕. It mimics a legitimate repository but serves malicious releases instead. I found it by searching the original repo in DDG. It only serves malicious executables, and it's been happening for more than 3 months.

@xpasky You could probably just use OpenRGB, which supports a lot of weird protocols and discovery.

> be me, buy a minipc sporting a shiny rgb stripe > can i turn it off? > found a windows binary blob that can control it, welp > download the zip and fire up pi > gpt-5.4: "reverse engineer LedControl_S3A_F3A.exe, i'd like to control the LEDs from Linux" (that's the full prompt) > 10 minutes of `objdump -d` later: "Yes — I dug into it." > i can now control my rgb stripe from linux > what.

@valyala @sasuke___420 @0xlelouch_ I have an article about similar approach rushter.com/blog/gzip-inde…

If you need fast location of any chunk in the file, then it is better to store chunk sizes and chunk offsets separately from the chunk data and name it "chunk headers". Chunk headers can be stored either in a separate file or at the end of the original file, so they can be appended there after all the chunks are stored to the file. If every chunk header occupies the same amounts of bytes (for example, 4 bytes per chunk size, plus 8 bytes per chunk offset), then finding the location of any chunk requires a single read from the chunk headers at the offset N*12 where N is the number of chunk to read.

If you have a 2TB gzip file, how do you parallelize processing when gzip is basically not splittable? What would you change in the storage format?

@max_spero_ @gwern Your Twitter bot gives some tweets 0%, but when you copy and paste it directly, it gives 100% generated. Something is not reliable.

No, you're right. This could be a failure of Pangram to properly capture the spectrum between AI-generated and human written. Our metrics are very good on truly human-written and fully AI-generated text, but text that has been co-written often evades Pangram entirely depending on the degree of human input. There are also adversarial prompting methods or paraphrasing attacks attacks, which can lower the percentage despite the original text being fully AI-generated. To me, it's hard to say which is which, but given the obvious errors it is quite possible that evading detection was the goal, and less effort was put into the post than we might otherwise believe.

thank you Gwern for repping Pangram on Hacker News. even if you called us "relatively easy to beat."

@ilin_pt 93% uptime, 65 incidents in the past 90 days

@rushter What’s so special about that chart?

@raysan5 This man can't stop

Today an AI-generated PR hit raylib repo, addressing a high-vulnerability security concern. 🔥 Interesting to see that it addresses a highly-improbable potential issue on a side tool that NOBODY actually uses, only me to generate some XML data for Notepad++ raylib intellisense. Oh, and **it BREAKS the PROGRAM** on my main working platform, Windows. Lovely. 😣 Definitely something to think about. 🤔 github.com/raysan5/raylib…

@ohmypy How about string and hash maps? Almost every project needs them, and when people use their own implementation, it's a source of bugs.

The only two features I really miss in C at this point are namespaces and defer.

@boscovne @LukasHozda Originally, they wanted to create a free Windows-compatible alternative. It gained a lot of traction when people were not happy with Vista and Windows 7, since it provided a familiar XP-like interface. Fifteen years later, it's still feature incomplete.

@rushter @LukasHozda What would the use case for this be? Keeping old -if-it-aint-broke-dont-fix-it systems running ?

With how badly is Microsoft treating Windows recently, the ReactOS team has an opportunity to do the funniest thing ever

@NearestCommit @LukasHozda They have a lot of goals, but the development is pretty slow. It's a hobby project. Just look at the release history: Dec 17, 2021 ReactOS 0.4.14 Mar 22, 2025 ReactOS 0.4.15 Release I've been reading about them since 2010. It was always pretty slow, but with big goals.

@rushter @LukasHozda The future goals of the OS is to also be compatible with newer versions of windows/NT

@roddux His point is that Proton can read your emails, not third parties

@wookash_podcast Every language has a preferred case

Why would language suggest a "proper" case?

@TheGingerBill Please add RSS

I've updated my blog's generator from using GoHugo to a very simple generator I wrote in Odin. ~630 LOC (at the time of this post) gingerbill.org/article/

@valigo Tufte CSS, web vs mobile

I think using footnotes in blog posts is a mistake - we are not in print media business, and there is a much better alternative. I really hope that people who write tech blogs consider it! (TLDW It's <details> tag)

Wrote a new article about ClickHouse internals that explain how it handles strings efficiently rushter.com/blog/clickhous…

@kerckhove_ts That usually happens when they remove spam messages that mention you. But it was fixed a few months ago.

GitHub regularly shows notifications that don't exist. Billion dollar company btw.

Wrote a new article about my Zsh setup rushter.com/blog/zsh-shell/