Sometimes I see people like 20 tweets deep into an argument with Grok. Like, what are you doing brother? You are trying to win an argument with a vending machine.

🎄 Introducing our paper A Generalizable Light Transport 3D Embedding for Global Illumination lnkd.in/gQUMSAyV . 🙈 Just as Transformers learn long-range relationships between words or pixels, our new paper shows they can also learn how light interacts and bounces around a 3D scene.

I may disagree with Sally Korbluth on many (most?) things, but credit where credit is due.

"Inside NVIDIA GPUs: Anatomy of high performance matmul kernels", includes a great intro to GPU architecture and PTX/SASS: www.aleksagordic.com/blog/matmul

Finally replied to one of these LLM generated undergrad emails saying they really liked my work on topic XYZ (I have never heard of this stuff let alone worked on it), and asked "What is topic XYZ? :)" I got back an LLM generated output explaining topic XYZ. Chat are we cooked?

I am boycotting #SIGGRAPHAsia 2026 because it will be in Malaysia, where there are serious safety concerns for the LGBTQ+ members of our community, thereby excluding them. I will not attend, submit my work to, or perform any reviews for SIGGRAPH Asia 2026. Yet, I'm excited about #SIGGRAPHAsia 2025!

Looking to talk to experts studying technology and race for 5 min today for a quick quote for something I'm working on. If this is you or if you know anyone who would be down, please reply or DM me! Hoping to chat ASAP.

Need for Speed Underground 2 website in 2004 #WebDesignHistory

Back when I was in undergrad, my mentor @hepcatjk.bsky.social used to talk about how one day someone was going to make the leap from pure particle based geometric optics to wave optics for graphics. That day has arrived.

what are we doing here people?

If you need to segment sketches into objects, Mia Tang has the tool for you. Enables re-composing the sketch after drawing. She presented her work on InkLayer at SIGGRAPH 2025. Code, data and paper are here: inklayer.github.io w/ @yaelvinker.bsky.social Chuan Yan, Lvmin Zhang

alternatively, replying to emails

Anyone here have experience PyPI packaging of custom nanobind bindings for CUDA code specifically? I'm fairly good at pybind/nanobind for standard C++ stuff, but I don't even know where to start for GPU code, especially one that links to Nvidia's proprietary libraries. pls halp 🙏🙏🙏

"Impractical in production" is my middle name ;)

My old rendering paper made it into the history of path guiding at the SIGGRAPH course! 😍

Ok but seriously now, why is SIGGRAPH North America never in Mexico? Seems like a fantastic option.

Meschers: Geometry Processing of Impossible Objects anadodik.github.io/publication/...

The Meschers talk is tomorrow! "Shapes, Surfaces & Forms" 3:45 – 5:35 PM West Building, Rooms 220-222!

Methods like NeRF and Gaussian Splats model the world as radioactive fog, rendered using alpha blending. This produces great results.. but are volumes the only way to get there?🤔 Our new SIGGRAPH'25 paper directly reconstructs surfaces without heuristics or regularizers.

Meschers got picked up by Gizmodo! gizmodo.com/meet-mescher... Thanks to @gayoung.bsky.social for the story!

The name is truly the biggest scientific contribution I've ever been involved in, stroke of brilliance on Kartik's part.

While I was in Rotterdam for CogSci '24, I visited the Escher Museum and fell in love with his work all over again. Now, a year later, I'm delighted by this SIGGRAPH paper led by my friend Ana! (Say what you will about the technical details, you must admit that we came up with the ~perfect~ title.)

Really excited to share this with you all! The website is already live at anadodik.github.io/publication/...

thank you Jim this is not demonic at all

I hope they make it tone down the em dashes so I can go back to overusing them without fear of people assuming I'm a slopper.

I really like this idea!

🥻Our new #SIGGRAPH2025 paper introduces a novel pipeline that captures and digitizes physical threads, predicts fabric appearance from the weaving pattern, and aids designers in exploring new possibilities. 🔗 mandyxia.com/research/fabrix.html

🤖 I really quite prefer this method of AI-enhanced answers to queries: Seeing sources alongside the highlighted portion determined to be directly relevant to your question. Ideally a few sources, esp. where there is multiple answers or disagreement. This combines multiple “AI” techniques.

Friends, we got a call to action 👇 We have the right to know when and how our creative works are used, especially when it comes to GenAi! This is why we need regulation like California’s Ai Copyright Transparency Act #AB-412 ! If you agree, especially if in CA, please sign & share this petition!

why is every shower user interface in the US completely unique? my airbnb has two twisty knobs on opposing walls, neither of which controls the water pressure. madness.

General public/news/congress/etc: gee , i wonder who we could talk to about making sure "AI" keeps the best interests of people mind Humanities ppl: Hi hello. GP/N/C/&c: oh, & socioeconomic implications! Gotta make sure we're working on that Social scientists: hello hi GP/N/C/&c: i know: Tech bros!

🚨🚨 WiGRAPH CONFERENCE COFFEE @ SIGGRAPH '25 🚨🚨 Sign up now to be randomly matched with peers for a SIGGRAPH conference coffee!

genius

System is so broken: - researchers write papers no one reads - reviewers don't have time to review, shamed to coauthors, use LLMs instead of reading - authors try to fool said LLMs with prompt injection - evaling researchers based on # of papers (no time to read) Dystopic.

Please stop assuming that too many em dashes means AI written. Some of us are producing grammatically incoherent work the honest way!

Meredith Whittaker is one of the smartest and most articulate leaders in tech I’ve ever seen, and I suspect that her thoughts on agentic AI and the like have more brain cells behind them than all of the rest of Silicon Valley combined.

baby's first demoscene demoparty

Guys it's finally happened. I went back and read a paper that's been sitting in my browser's tab bar for months.

New paper out today, accepted at #FAccT2025, led by Jake Stein and me: "Not Even Nice Work If You Can Get It; A Longitudinal Study of Uber's Algorithmic Pay and Pricing" arxiv.org/abs/2506.15278 Over the last 3 years, we've been working with @workerinfox.bsky.social to audit Uber's algorithms. 1/

Credit to @geometry.gay and Isabella Yu for the GDP mug, whose letters are “impossible shapes” designed using our Meschers tool; and to SGI 2025 Fellow Marina Levay for the secret-geometric-message SGI mug.

New MIT GDP Group (left) and Summer Geometry Initiative (right) mugs just dropped! Collect them all: visitors to our lab get GDP mugs, and SGI mentors/volunteers/participants get SGI mugs. ☕️🫖🍵

New research from MIT found that those who used ChatGPT can’t remember any of the content of their essays. Key takeaway: the product doesn’t suffer, but the process does. And when it comes to essays, the process *is* how they learn. arxiv.org/pdf/2506.088...

Mesh intersections is hard ! Three years of agonizing pain summarized in my article that just got accepted to ACM Transactions on Graphics dl.acm.org/doi/10.1145/... The implementation and the expansion-based arithmetic kernel are available in geogram: github.com/BrunoLevy/ge...

MIT News today features a paper with fun examples coming out this summer by our brilliant PhD student Leticia Mattos Da Silva, postdoc—and Columbia professor in a matter of days!— @silviasellan.bsky.social, and PhD student collaborator Natalia Pacheco-Tallaj. news.mit.edu/2025/animati... Flubber!

Ever wondered how badly we're all addicted to buying new GPUs in graphics labs? Come see our talk at #SIGGRAPH2025 to discuss how we can collectively move "Towards a sustainable use of GPUs in Graphics Research" with @elie-michel.bsky.social @axelparis.bsky.social Octave Crespel and Felix Hähnlein

Long live the triangle!

Thank you to everyone speaking up about this publicly. I am unsure why it had to come to this, but I sincerely hope that SIGGRAPH Asia will reverse course. alecjacobson.com/weblog/siggr...

Check out DiffLocks, appearing at #CVPR2025. From a single image, we estimate about 100K hair strands that you can then physically simulate. We use a dataset of 40K synthetic hair images with ground truth strands. It's all available for research purposes.

Me with textbooks.