Senior Researcher @ intrinsic.ai (AI Robotics group at Google). Here talking about 3D Computer Vision, Robotics, and life. Views are my own.
San Francisco
- Tweets 1,150
- Following 500
- Followers 376
- Likes 1,122
Pinned Tweet
Placing a risky bet and watching it pay off is one of the best feelings in life.
A year ago, we set out to achieve SoTA in stereo 3D reconstruction. Instead of building a complex model, we did the opposite: we stripped it down to a simple ViT.
Introducing NBS 🧵👇
20
21
207
13,790
We need more postmortems like this one. Very honest and nicely written. I was concerned about the robotics data market already, this has confirmed some of my concerns
After 2+ years in the robotics data space, we are shutting @Eidon_AI down.
The thesis was right. But the business is brutally hard.
We close this chapter by open-sourcing everything we built and sharing lessons for anyone venturing into the space.
1
132
Vahe Taamazyan retweeted
Workshop on Recovering 6D Object Pose, this Tuesday at ECCV'26 in Malmö!
🗓️ 14:00 - 19:00 on September 8 in Quality View Hotel (room Stroget 1).
🤖 This year we focus on 3D/2D object grounding and robotics.
🕸️ Website: cmp.felk.cvut.cz/sixd/worksh…
💡 Invited talks:
@RanjayKrishna (@UW, @MicrosoftAI) - It Is Time to Revisit Grounding
@b_k_bucher (@UMich) - How Much Pose Does a Robot Need? Lessons from Mobile Manipulation
@ZakharovSergeyN (@ToyotaResearch) - Closing the Loop: Synthetic Worlds for Scalable Robot Learning
Ian Chen, Trushant Adeshara (@IntrinsicAI): AI for Industry Challenge
@tomhodan (@MistralAI ), @ma_sundermeyer (@Google) - BOP Challenge 2026 & BOP-Refer Benchmark
See you on Tuesday!
4
9
31
61,593
These meme from @Parskatt is even more relevant with NBS
Placing a risky bet and watching it pay off is one of the best feelings in life.
A year ago, we set out to achieve SoTA in stereo 3D reconstruction. Instead of building a complex model, we did the opposite: we stripped it down to a simple ViT.
Introducing NBS 🧵👇
5
802
Placing a risky bet and watching it pay off is one of the best feelings in life.
A year ago, we set out to achieve SoTA in stereo 3D reconstruction. Instead of building a complex model, we did the opposite: we stripped it down to a simple ViT.
Introducing NBS 🧵👇
20
21
207
13,790
Meet NBS: SoTA stereo model (ranked #1 on ETH3D) with zero stereo inductive bias and 5-10x faster inference than FoundationStereo and S2M2.
1
18
1,425
We hope NBS pushes the frontier for 3D vision and inspires more research into bias-free, highly scalable stereo models.
Check out the project page here: intrinsic.ai/publications/NB…
20
1,380
Stolen from memegen
Just shared some changes we’re making to the teams at @GoogleDeepMind.
@DemisHassabis is stepping up to become Chair of @GoogleDeepMind & Chief Scientist of Alphabet, in addition to leading @IsomorphicLabs. He’ll be able to dedicate his time and focus on shaping the future of AGI and scientific discovery. It’s work that is vitally important to Alphabet and humanity, and I can’t imagine a better person than Demis to do it. He’ll stay closely connected to Koray and the GDM teams.
@Koraykv will become the SVP, @GoogleDeepMind, responsible for all aspects of model development, GDM research, and @Geminiapp & dev teams. Koray has been at GDM for 13 years and is a world-renowned expert in the field, starting our deep learning team and driving breakthroughs like WaveNet & DQN. GDM is in great hands!
Excited for this next chapter. You can read my note along with the message Demis sent to @GoogleDeepMind here: blog.google/company-news/ins…
3
212
Your daily reminder that the development of advanced AI is not the end, its just the beginning. We have so many unanswered questions and mysteries about the universe around us that AI can help us with
2
141
Has anyone posted the full reasoning trace of the models coming up with counterexamples to conjectures? What is the sequence of thoughts and ideas that leads them to the result? I feel like this is the most important thing for us to understand
96
My every workday for years has started the same way: open arXiv's CV page, scroll 200+ new titles, hunt for the 2–3 papers I actually need. 20–30 minutes a day. I haven't skipped a day even on vacation.
Last month I finally made an LLM take over the ritual.
1
192
Every morning the bot reads every new CV paper, asks Gemini one question per paper - "is this relevant to his interests?" - and emails me a digest of the survivors.
Weeks of testing so far: zero missed papers, and my reading list is ~80% shorter.
1
58
The whole thing runs completely free: Gemini's free tier + free email + a cron job on any computer you have lying around.
It's open source. Swap in your own research interests and let it read arXiv for you: github.com/vaheta/arxivbot
1
1
47
A side project post:
@petrkaplunovich and I built a free meditation website a while ago - and we never told anyone.
It's inspired by a NYT story about a Harvard art historian who has her students spend 3 hours looking at a single artwork.
👇
1
2
3
139
Every day there's one new piece from the Met's open-access collection, and nothing else.
Honestly, everyone on X needs to use this to chill from all the AI FOMO and culture wars. Just look at some beautiful artworks folks!
lookatarts.com
54
Vahe Taamazyan retweeted
📢 11th R6D workshop on Sep 8 at ECCV'26 in Malmö!
cmp.felk.cvut.cz/sixd/worksh…
🛝 This year focused on visual grounding capabilities of vision-language models (VLMs)
📊 New BOP-Refer benchmark on 2D/3D grounding
📄 Paper submission deadline: July 24
3
6
35
3,772
Since when is @homescreen__ newsletter just another random AI slop with factually wrong info? The Gemini 2.5 Pro release was like a year ago.
1
232
Why isn’t there a car maker offering a KITT level intelligence and functionality in the car? Tesla seems to be the closest one, but it doesn’t seem to be as fun. The tech is already at the level. I’d love one
1
84
If you get one day with complete understanding of the universe - why it exists, everything - but it’s your last day alive, would you take it?
1
75
