Social Animus

May 28th, 2026 @ justine's web page

The most difficult challenge to working in open source is that there's no institutional screening process, since the goal is to just let people organize themselves and build things. This has meant that many of the people who get involved have never had the opportunity to work with the most exemplary members of each group the world has to offer. During the culture wars of the 2010s, the first person who tried to solve the problem of how to include these uncommon individuals was Coraline Ada Ehmke, who wrote the Contributor Covenant. I always thought her solution went too far, since I found a much easier answer for my own project, which has been to never accept anonymous contributions and to not merge a single line of code until the contributor sends an email promising to assign me copyright.

I always thought my security posture was too paranoid, so when llama.cpp came out in 2023, I found the code Gerganov wrote to be so beautiful that I did the one thing that I promised myself I would never do, which was collaborate with an anonymous developer from his team named Slaren. This was the first time in five years that I wrote a change with someone on a project that wasn't my own. After submitting our work he went on 4chan afterwards and accused me plagiarism, saying that even my changes were his own. The way the community reacted is an interesting case study into the guile some developers have learned since the culture war, because the locus of thought for llama.cpp has always been on 4chan. They were the ones who originally leaked the Meta LLaMA v1 weights. You can map the way developers talk on that board to their anonymous accounts on GitHub. I actually developed migraines for the first time in my life and ended up in the hospital (since I didn't have health insurance and had to wait in the ER) due to the eye strain of reading unfiltered thoughts about me for months.

It's unusual because the community originally reacted positively towards my work, until one of its members felt threatened by me, and since they're all anonymous there's not much proof it wasn't just a few guys. This was the reason Wendy Hanamura cited when she canceled my invitation to speak at the Internet Archive.

In any case, I'm really happy that these back channels exist, because the greatest competitive advantage I've ever had was to monitor which pull requests people on 4chan complained about, and then merge them into llamafile before Gerganov could. This is how my Mozilla Builders project shipped support for new models like Gemma 2 before any other grassroots project. I got hundreds of thousands of downloads on Hugging Face. There were so many downloads that Mozilla couldn't believe it, because so few people showed up on our issue tracker. Mozilla was sponsoring my work because they want to support the community, and as far as anyone could tell, there wasn't one. I always thought this happened because my code was just that good. In a past life, when I was originally trained to write kiosk software for reverse vending machines in Java, no one ever contacted the vendor unless there was something wrong, and since llamafile is an ex nihilo project that I worked on for six years, beginning with an empty file and an assembler, I had plenty of time to pin down most of the bugs on my own.

I even wrote a blog post giving Slaren more credit, because it instilled in him a false sense of confidence that led him to tackle harder problems, like multiplying three dimensional numbers. To fix the performance issues with mixture of experts models that caused, I tried to upstream my tinyBLAS tensor multiplication code in PR #6840 and it's a great example of what it's like to work with me. Gerganov's doctoral advisor was Iwan Kawrakow, who was the power behind the throne on that project. He invented the "K" quantization formats many people use to compress their weights. He was curious about my change and I told him that he'd be able to build better matrix multiplication kernels than me if he used my block tiling technique with his quants.

llamafile ended up receiving an avalanche of pull requests from Iwan that were licensed Apache 2.0 so that Gerganov couldn't use them. This enabled us to have faster cpu inference than any other project. That meant consumers and businesses stood a better chance of being able to use LLMs without needing to purchase expensive GPUs. We made that happen, even though the llama.cpp team had more than a million dollars of funding, and were successfully acquired by Hugging Face after Iwan had moved on to start his own project.

Hacker News is my favorite place on the web, because it's the last bastion of curiosity online. This was the cheat code I used to restart my career in 2020. The first thing I did was I wrote an article about αcτµαlly pδrταblε εxεcµταblε and I hosted it in a Google Cloud Storage bucket. When users voiced concerns about the optics of my CDN URL and its use of the Greek alphabet, the Hacker News moderator dang said:

People have of course been pulling such tricks for years. We don't allow it generally, and in fact are pretty strict about that. But all such rules have exceptions. The quality of this post is so high that it doesn't feel right to override any aspect of what the author created, including quirks like the title. There may be a superficial similarity to garden-variety title pimping, but as someone who bathes in spam every day I can tell you how rarely such gimmicks come attached to first-class feats of hacking prowess. Respect for content requires looking closely enough to detect truly unusual cases, pick them out of the mud, and give them a special place.

I was so happy to be welcomed that day, that I went on to build other hobby projects like redbean, which is the sixth most upvoted Show HN thread of all time. I also made SectorLISP (the world's tiniest LISP interpreter) and SectorLAMBDA (a binary lambda calculus interpreter in a 383 byte x86-64 executable).

After a few years of doing this, doors finally started opening up for me. Mozilla asked me to come to MIECO meetings each week. One of the happiest days in my life was when a Wikipedia editor came along and cleaned up the BLP issues in my article. I was so glad to be treated fairly by a great website, even if I only have a short article, because most of the work that went into my biography was adjudicating concerns about whether I should be included at all.

llamafile is based on my Cosmopolitan C Library and LWN wrote an article about it that explored its technical depths better than anything that'd been written previously, including by me. The site's owner Jonathan Corbet is easy to influence. He doesn't moderate his website and he limits his admin powers to avoiding conflict. When Jürgen Geuter told him that cosmopolitan is an antisemitic dog whistle and that my existence is about destroying democracy, Jonathan decided to add an apology to the top of the article while continuing to host the bad faith criticism and then prevented new comments from being made.

Lobsters is the best place to go when you want candid feedback. My AI Training Shouldn't Erase Authorship blog post originally had a more emphatic title. One of the issues I have is that I'm so popular on Hacker News that people there don't criticize me so much these days, even when I'm wrong. But since you can always count on Lobsters to tell you what they're thinking, when I saw someone complain about what he thought the title should be, I just used his idea instead of my own. This upset the moderator so much when he saw that I was optimizing my writing style based on feedback from his site, because he had already banned me for spamming, along with every single person I invited too, like woodrush, whose LAMBDALISP project I encouraged him to share. So this became the day Lobsters also banned my domain, so that no one else could post my articles.

The U.S. government comes across today as not wanting people like me to participate in public life.

Since I'm a volunteer, and as a general rule in life, I only try to help others when my help is wanted and needed. So when the Don started his second term, I get a job at the Gradient Canopy to help improve the performance of their tensor processing units for Gemini. It's a great place to be invisible since it's a building so secretive and so secure that even googlers aren't allowed to just waltz in and grab a free lunch. Inside you'll find Sergey surrounded by developers who put more thought into brewing their coffee than Gale Boetticher, working on a codebase that's filled with Einstein summations. I walked and rode those little gbikes so much that I lost twenty pounds. It's basically like being exiled to Heaven. Unfortunately the friend who fought hard for me to be included had to go on paternity leave. Then the company spent billions of dollars to hire Noam Shazeer, who's listed as an author of the Attention is All You Need paper so folks think he invented AI. The first thing he did with his enormous wealth and power is share his animus towards trans people and I got fired for performance reasons around that time.

All three of the people I worked with at Mozilla ended up moving on to different things after I got curious about TPUs, because we thought the llamafile project failed. But Mozilla still managed to find a team that loved it enough to keep working on it. One of their goals has been to introduce support for new models. They decided to do that by unvendoring the project and synchronizing it with the canonical upstream code. They wrote a great report about the 15x performance regression in MoE models that caused. That's because they're still working out how to bring back Iwan's code, Gerganov stopped merging my changes too, and he didn't have room for GeLU.

I want the llamafile team to move fast and break things, because they're independently verifying what I can't easily explain. Since llamafiles use my Actually Portable Executable file format, the release artifacts are engineered to last forever.