MTS @VinciPhysics, founding team @ uber ai labs & @ml_collective & geometric intelligence, ex toy designer @KiteandRocket 🌈✡️
San Francisco, CA
Joined March 2012
- Tweets1.4K
- Following3.9K
- Followers1K
- Likes69.6K
Eric Frank retweeted
Argon has this amazing natural understanding of the world. It can do so many things from first principles.
Here's a pelican riding a bike, using fourier epicycles.
Video made by Argon.
Time-synced Lyria 3.5 music prompted by Argon.
Gemini 4 Argon is live.
blog.google/innovation-and-a…
There's a lot to show you.
Eric Frank retweeted
$250M in new funding. A $1.5B valuation. A conviction that the way we engineer the physical world is about to change.
Today, Vinci announces our Series B, co-led by @adventintl , @Temasek and Xora Innovation, with participation from @AMD Ventures, @EclipseVentures , @khoslaventures, @MadronaVentures and others.
This investment backs a future where physics intelligence becomes foundational to how hardware is designed and built.
For decades, engineers have rationed analysis, narrowed exploration and committed to decisions before knowing the physical consequences. Vinci’s Continuous Physics Reasoning makes deterministic, solver-accurate physical understanding continuously available as designs evolve. It is already running on more than 40 flagship production engineering programs.
This round fuels our next chapter: expanding across more physics and more of the hardware economy, moving from understanding what will happen to helping engineers determine what should change.
Semiconductors are the proving ground. The ambition reaches everything we build.
Thank you to our customers, team and investors for making this possible. The next era of hardware engineering is taking shape. We’re building for it.
Eric Frank retweeted
BTW, the entire development vault of Esperanto is up for open-sourcing/public-release, need to find funding to pay an intern to walk through the files and make sure no proprietary TSMC stuff is leaked
Eric Frank retweeted
Spent last two weeks learning electronics by having Opus run experiments on the Esperanto AI chips and condensing the results into interactive diagrams like this spacesheep.dev/@yaroslavvb/e…
Eric Frank retweeted
Ive slowly came to realization that more and more people hate AI. Artists hate it, programmers hate it, mathematicians hate it, and everyone thats currently impacted in business hate it. It used to be small number of people, but its quite a lot now.
I've been truely in a bubble because for past one year ive just met my peers in AI field / my workplace and my close friend, and noone else outside my existing network group.
Ai researcher is slowly becoming most hated occupation and I dont know what / how to feel about it. We now have some big obligation to 'make it all work', i.e., permanent upperclass for everyone, or get hanged.
Eric Frank retweeted
I want a futurism that feels like this. Like we've mastered nature and mastered the human-scale longed for ideals.
Not like flying plastic or plants on top of glass-plastic
Eric Frank retweeted
AIs cracking short-term superforecasting is such an important (and, still, under-discussed) trend.
Via a new piece from @nikostro economist.com/science-and-te…
Eric Frank retweeted
Stop blowing up Looped Transformers — it's not magical, it doesn't add recurrence to the (still and always) feedforward network, it was done years ago (Dehghani et al., 2019). All it does is making your model deeper, the same as if you added more layers to begin with.
Eric Frank retweeted
watch all the accounts rn who have absolutely no clue what the difference between "model architecture" (latent space manipulation) vs "thinking traces" (model output) actually are
Eric Frank retweeted
New #preprint - @YanboZhang3
"Intelligence from Learnable Novelty"
arxiv.org/abs/2607.18433
What if we optimize Epiplexity (arxiv.org/abs/2601.03220 @m_finzi @andrewgwils ) instead of measuring it? We have derived a closed-form approximation of Epiplexity and discovered a deep connection between it and intelligence. This allows us to reinterpret Epiplexity as a form of learnable novelty, providing a brand-new understanding of what intelligence is. By maximizing Epiplexity across various systems, all of them exhibited interesting behaviors:
Cellular Automata: Maximizing Epiplexity directly generates complex soliton interactions similar to Rule 110.
Image Encoders: It automatically causes the encoding to cluster, successfully categorizing different handwritten digits without supervision.
Reinforcement Learning: Introducing Epiplexity improves the performance of PPO in sparse reward tasks.
We also explored the relationship between the theory of learnable novelty, the free energy principle, and novelty search. We hope this work helps us better understand the nature of intelligence and its origins.
Eric Frank retweeted
Glad to share that AdaSplash-2 was accepted to ICML 2026! 🇰🇷
Btw, we've released the Triton kernels for AdaSplash-2 in our GitHub repo. Nuno's histogram implementation with bit-packing to avoid atomics is, in my totally unbiased opinion, remarkable!
github.com/deep-spin/adaspla…
Eric Frank retweeted
Why aren’t Diffusion Language Model smart yet? Lacking stable post training is a major bottleneck!
Meet DiPOD: the tripod for diffusion model post-training.
DiPOD boosts accuracy across reasoning tasks, with Sudoku jumping from 22% to 97%, through a one-line code change.
🧵1/5
Eric Frank retweeted
Neural PDE solvers have seen exciting progress! 🌊
But despite growing adoption, we still don’t know 𝘄𝗵𝗲𝗻 we should use them instead of classical solvers. 🤔
Our new paper has a surprising finding: the harder the PDE task, the more cost-effective learned solvers become. 🧵👇
Eric Frank retweeted
We never really knew how to train nonlinear RNNs well… BPTT struggled with vanishing grads (no long-range memory) and sequential rollout (hard to parallelizable).
What if instead an oracle told us the optimal memory state m_t at each step? Then the RNN could do one-step supervised learning on (m_t, x_{t+1}) → m_{t+1} labels.
We call this Supervised Memory Training (SMT): a replacement for BPTT that trains RNNs without unrolling them. SMT is time-parallelizable and solves vanishing gradients.
Website: akarshkumar.com/smt/
arXiv: arxiv.org/abs/2606.06479
Eric Frank retweeted
With UMA playground, the crazy things Feynman could only ask us to imagine become something you can see. Heat, break, and build at aidemos.atmeta.com/uma
dm me if you’re at cvpr and interested in scaling physics ai models, procedural content generation or active learning
Eric Frank retweeted
Most engineers have never had sufficient access to real physics analysis — simulation was too expensive, slow, and specialized. @VinciPhysics vision isn’t just replacing Ansys; it’s bringing continuous physics reasoning to 100x more engineers and 1000x more simulations in a fraction fof the time.
Every major AI shift started with a new kind of world model.
Language. Vision. Code.
The next one is physics.
@saucentoss and I published a paper today defining what a real foundation model for physics has to be and why it enables Continuous Physics Reasoning.
If AI is going to help build the physical world, the bar has to be much higher.
Link to full paper: vinci4ddev.wpenginepowered.c…