Terence Tao is on every good podcast right now and I’m not complaining.
Terence Tao is arguably the greatest mathematician of his generation, but Tyler spent this episode asking him not what the proofs say, but what they feel like.
Some of Tyler’s questions for Terence in this episode, ranging from what emotions Euler’s identity and the Riemann hypothesis evoke, to whether AI has ended the era of individual open problems in math, to whether the Collatz conjecture is actually true:
· What does Euler’s identity feel like to you?
· What does pondering quantum mechanics feel like?
· What do you feel when you think about the Riemann hypothesis?
· Does Gödel’s incompleteness theorem frustrate you?
· How much do you read in the history of mathematics, and why?
· Is most abstract math useful, or a lottery ticket?
· Does string theory feel beautiful to you, or just strange?
· Why did real analysis arrive so late in mathematical history?
· Why did it take until the 7th century to invent zero?
· Were you shocked that probability doesn’t matter in Black-Scholes?
· Are we living in a simulation?
· Are you a one-boxer or a two-boxer on Newcomb’s paradox?
· How would you fix math education in America?
· Has AI pushed math from proof scarcity to proof abundance?
· Will the research paper survive as the basic unit of mathematics?
· What’s your favorite Hong Kong action movie?
· Would you have done worse at Cambridge or Princeton instead of Flinders?
· What’s your most unusual successful work habit?
· Is the Collatz conjecture true?
Watch the full conversation with Terence Tao and @tylercowen here, or search for Conversations with Tyler wherever you get your podcasts.
youtu.be/XB5oudgNWO8?si=Rb8G…
28
Meet Déjà View, a model that turns photos taken from different angles into 3D geometry and camera positions.
By looping a single transformer block, this 117M-parameter model matches or outperforms models up to 10x larger across five benchmarks, using comparable or less compute. Researchers can dial the number of loops to fit their compute budget.
#NeurIPS2026 oral paper + code: research.nvidia.com/labs/dvl…
Bruno Santos🇵🇹 retweeted
To match or not to match?
FlexGlue 🐙, our new image-matching model, can match every pixel AND reject doppelganger pairs, tackling a common failure case in SfM.
➡️State-of-the-art SfM and matching, blazing fast
➡️Code: github.com/cvg/FlexGlue
➡️Paper: arxiv.org/pdf/2610.32976
1/
Bruno Santos🇵🇹 retweeted
Robotics is solved when world GDP has grown by a hundred orders of magnitude and we've collonized the galaxy, not when SOTA on some benchmark goes from 42.2% to 63.1%
Bruno Santos🇵🇹 retweeted
A ‘hello world’ milestone for Project Suncatcher 🚀
After launching our prototype satellite a few weeks ago, we’ve now successfully sent a command up to orbit, booted up the TPU, ran the prompt, and received our first complete response back on earth - a haiku!
Bruno Santos🇵🇹 retweeted
Introducing Echo-3. A state-of-the-art world model.
Echo-3 turns phone captures into photorealistic, metric-scale 3D worlds. From a single room to multi-floor buildings, outdoor and public spaces.
Available today.
Mathematician Terence Tao:
"we're seeing the end of the romantic era of mathematics, the age of heroes. we're moving into an industrial era where math becomes large-scale, done in big teams like the other sciences"
We're entering, rather unwillingly, a new era where exciting new things are possible, but some things are being lost
Bruno Santos🇵🇹 retweeted
CF-Splat: Closed-Form Triangle Splatting with Mesh Priors
Yixin Fang, Liang Zhao
tl;dr: mesh geometry&calibrated cameras->rendered pixel colors are linear with respect to per-vertex SH coefficients->multi-view appearance estimation->global LLS problem
arxiv.org/abs/2610.22108
Bruno Santos🇵🇹 retweeted
DINOv2 has never seen a caption, and Qwen3 has never seen an image. We still aligned their embedding spaces without a single image-caption pair.
It even works when the images and the captions come from different datasets.
Project page: dominik-schnaus.github.io/un…
⬇️
Great talks on egocentric vision .
You can now watch the entire Fifth Hands-On Egocentric Research Tutorial with Project Aria from Meta from ECCV 2026!
11 talks from the brilliant minds behind Boxer, LAMP, Eye Tracking on Aria Gen 2, plus some amazing partner work from University of Maryland, ETH Zurich, TUM, and more. 💡
youtube.com/playlist?list=PL…
3
321
The story plot about reaching professors in academia looks pretty like a soap opera .
E-mails are overrated.
LinkedIn DMs underrated .
1
176
SITUATION DETECTED: OpenAI has released a GitHub repository with over 700 math problems solved by its models.
github.com/openai/math
Bruno Santos🇵🇹 retweeted
Clone is the humanoid company closest to the *Westworld* vibe,but it has redesigned its entire system.
Clone just showed teleoperation of the hand of Torso 3--VR-driven, matching open-close, point, and pinch motions.
The founder,Dhanushi says this hand was built in May,and is the last of their legacy hand designs.
They’ve spent the past six months rebuilding the whole system from the ground up, aiming for better simulatability and controllability without giving up near-human capability.
That work lands as Torso 4 in the first week of November--a torso for fixed, two-handed enterprise tasks.
It is worth noting that-- no gears anywhere in the system.
Here, we must mention the Protoclone, the android prototype launched by the Polish company Clone Robotics in February of this year.
Protoclone features a polymer skeleton that mimics the 206 bones of a human.
Encasing the skeleton is a humanoid artificial muscle system: equipped with over 1,000 artificial muscles based on the company's "Myofiber" technology, similar to the concept of McKibbin pneumatic muscles.
These muscles mimic the function of human muscles by contracting balloons within a mesh tube filled with hydraulic fluid.
























