Jon Barron

2.8K posts

Jon Barron

@jon_barron

Principal research scientist at Google DeepMind. Synthesized views are my own.

SF Bay Area Katılım Mayıs 2010

1.3K Takip Edilen33K Takipçiler

Sabitlenmiş Tweet

Jon Barron@jon_barron·28 Nis

Here's my 3DV talk, in chapters: 1) Intro / NeRF boilerplate. 2) Recent reconstruction work. 3) Recent generative work. 4) Radiance fields as a field. 5) Why generative video has bitter-lessoned 3D. 6) Why generative video hasn't bitter-lessoned 3D. 5 & 6 are my favorites.

English

102

787

113.4K

Jon Barron@jon_barron·1d

Sam Roweis was my undergrad advisor, and is very regrettably deceased. He had a lovely PDF for Gaussian identities that I've used for years: people.orie.cornell.edu/andrew/orie674… I've turned it into a markdown file for posterity: github.com/jonbarron/gaus… RIP Sam, thanks for everything.

English

296

26.7K

Jon Barron retweetledi

Marc Habermann@marc_habermann·2d

🔥 #3DV2026 Oral: We present a matrix-free Levenberg-Marquardt optimizer that makes second-order optimization practical for 3DGS! 🚀

English

334

24.6K

Jon Barron@jon_barron·2d

@chrisoffner3d Video game people tend to not realize that they've gotten used to (or even desire) an aesthetic that most people without exposure to games find strange or even disturbing. I think it's okay fwiw, "2010-era video game" is just its own aesthetic now, hopefully one of many to come.

English

906

Chris Offner@chrisoffner3d·3d

I’m really surprised how angry gamers online™️ are about DLSS 5 and how many actually claim that the gamey graphics on the left look better than the AI graphics on the right. You’re just more used to the artifacts and limitations of the left tech than the right but come on!

kron123456789@kroshkaruiya1

@compusemble Well yes, but actually no. I mean, yes, it may look more photorealistic, but I really struggle to call it better

English

5.1K

Jon Barron@jon_barron·2d

@GS05445168 True, but on the other hand: dishwashers, etc

English

532

Yottachad@GS05445168·2d

@jon_barron There needs to be an at home robotics deployment that does not involve turning your home into a distribution center

English

696

Jon Barron@jon_barron·3d

There's an engineer on YouTube building his own room-scale laundry-picking UFO catcher robot out of QR codes and string, it's one of the most compelling robotics demos I've seen in a while.

English

244

507.7K

Jon Barron@jon_barron·2d

@nickcammarata I think we're just seeing a brute force search or a random walk, but our brains seek out an explanation that has intentionality behind it because that's what we always do.

English

933

Nick@nickcammarata·3d

I don’t actually know how to process this, what’s a reasonable explanation of what is going on here

The Figen@TheFigen_

They are ants solving a geometric problem and it is mind-blowingly colorful.

English

210

1.4K

270.6K

Jon Barron retweetledi

Junyoung Seo@jyseo_cv·2d

What if a world model could render not an imagined place, but the actual city? We introduce Seoul World Model, the first world simulation model grounded in a real-world metropolis. TL;DR: We made a world model RAG over millions of street-views. proj: seoul-world-model.github.io

English

206

1.5K

164.5K

Jon Barron@jon_barron·3d

@cjbarnaby Yeah a big part of technology is solving a hard problem once so that you don't need to solve an easy problem a million times

English

2.5K

CeeJay Barnaby@cjbarnaby·3d

@jon_barron so lazy he can't pick up his dirty clothes but so driven to make a robot for that. weird.

English

Jon Barron@jon_barron·3d

@bilawalsidhu The boost in image quality really draws attention to the animation's lack of realism, I think this takes us slightly deeper into the valley overall. Gotta get all the way across it.

English

2.3K

Bilawal Sidhu@bilawalsidhu·3d

Big moment. You can cross the uncanny valley in video games by using real-time video-to-video AI. You get the best of coherence & control from classical 3d engines, then use generative AI to take it all the way.

NVIDIA GeForce@NVIDIAGeForce

Announcing NVIDIA DLSS 5, an AI-powered breakthrough in visual fidelity for games, coming this fall. DLSS 5 infuses pixels with photorealistic lighting and materials, bridging the gap between rendering and reality. Learn More → nvidia.com/en-us/geforce/…

English

204

32.3K

Jon Barron@jon_barron·3d

@AverageNBAFannn this seems easier

English

AverageNBAFan@AverageNBAFannn·3d

@jon_barron Just pick up your laundry bro

English

Jon Barron@jon_barron·3d

@cmarschnerde oh TIL. I'm used to calling them "fiducials" which has always felt awkward.

English

Clemens Marschner@cmarschnerde·3d

@jon_barron These are ArUco markers, an easy way to localize something in a picture

English

4.3K

Jon Barron@jon_barron·3d

His channel: @overengineer7691" target="_blank" rel="nofollow noopener">youtube.com/@overengineer7…

English

12K

Jon Barron@jon_barron·4d

@Azadux The tone mapping on the asset is slightly punchier than the passthrough, not sure how to fix it though

English

1.4K

Azad Balabanian@Azadux·4d

Woah, Gaussian splats in mixed reality look so good! Tech stack: lichtfeld studio for splattering training, virtual desktop for PC VR streaming and passthrough MR using a virtual green screen, gracia for VR splat rendering using the green screen background feature

English

367

36.2K

Jon Barron retweetledi

dmvrg@dmvrg·6d

@justinryanio On Quest, you can already switch to a fully immersive MR mode (there is no 'Launch in VR' button on Vision Pro yet)

English

3.9K

Jon Barron retweetledi

Xuan Luo@XuanLuo14·6d

📢 Google's Project Beam (formerly Project Starline) is looking for Student Researchers! We seek students interested in real-time high-realism 3D reconstruction, dynamic 3D, relighting, materials, novel view synthesis, diffusion, video models, transformers, and more. 🧵1/7

English

205

23.2K

Jon Barron@jon_barron·6d

@Andercot I expect skill markdown content will just migrate to docstrings. Even if the same content exists in the code, compressing it into less text and making it easier to find will save the agent compute no matter how smart they get.

English

825

Andrew Côté@Andercot·12 Mar

I give it about 3-6 months before any kind of skills.md file is also pointless. The same thing happened to vector databases and langchain and every other 'product' built in the narrowing gap of model competencies.

@levelsio@levelsio

Thank god MCP is dead Just as useless of an idea as LLMs.txt was It's all dumb abstractions that AI doesn't need because AI's are as smart as humans so they can just use what was already there which is APIs

English

166

1.5K

414.9K

Jon Barron retweetledi

Michael Niemeyer@Mi_Niemeyer·12 Mar

Hiring: Student Researcher @ Google Zurich! 🇨🇭 Looking for an EMEA-based PhD student to work on: Efficient 3DGS / Feedforward Models / 3D GenAI. Start: June 2026 or earlier. If interested, please reach out with your CV to: sr-zurich-3d@google.com

English

482

42.7K

Jon Barron@jon_barron·12 Mar

@ducha_aiki I'm still not totally sure why that is a rule

English

1.9K

Dmytro Mishkin 🇺🇦@ducha_aiki·12 Mar

@jon_barron You still would need papers to get a PhD if you care about degree. The rest is good plan

English

2.2K

Jon Barron@jon_barron·10 Mar

If I was a grad student today, I would: 1) Not write papers, 2) push my (agent-written) code to a public repo ~weekly, 3) maintain (via agents) a writeup.tex (manually verified) and a skill.md in the repo, and 4) work towards establishing skill usage as the new "citation" format.

English

574

92.9K

Jon Barron retweetledi