Juspreet Singh Sandhu (@JuspreetS) header
Juspreet Singh Sandhu (@JuspreetS) profile picture

I like TCS/Mathematical Physics, Cycling, Classical Music, Chess, Photography & Korean/Indian Food.

Fort Collins, CO
Joined December 2020
Photo by Juspreet Singh Sandhu (@JuspreetS)Photo by Juspreet Singh Sandhu (@JuspreetS)Photo by Juspreet Singh Sandhu (@JuspreetS)Photo by Juspreet Singh Sandhu (@JuspreetS)Photo by Juspreet Singh Sandhu (@JuspreetS)Photo by Juspreet Singh Sandhu (@JuspreetS)
Building on the spin-glass & free-probability estimates we developed in PHA 3 @oldheneel, @jtnshi and I observed that the "regularity" of the ASL process + conductance can be used to show WPI for a target Gibbs. The framework is applicable under mild regularity conditions 1/2
Weak Poincaré Inequalities via Approximate Stochastic Localization: Application to Sampling the Sherrington-Kirkpatrick Model Ewan Davies, Holden Lee, Juspreet Singh Sandhu, Jonathan Shi arxiv.org/abs/2607.08160 [𝚖𝚊𝚝𝚑.𝙿𝚁 𝚖𝚊𝚝𝚑-𝚙𝚑 𝚖𝚊𝚝𝚑.𝙵𝙰]
1
1
1
13
7,755
It is nice to see there are a few who understand what the true spirit of mathematics is (in my view, at least). This is the sort of ownership a craft like this demands. I wish the author the very best, and hope we all gain understanding and more questions & ideas from the work!
A statement from Alessandro Masullo about his proposed proof of RH. He didn't plan to post so soon and knows there is work to be done. "A human being has come up with these ideas. A human being has been assembling this math. I gave my word that I would see through through, all the way. This is me keeping it."
2
15
756
This is how adults work. They patiently get experts to review what they've generated, share human understanding, allow their work to be stress-tested and focus on exposition. They don't just lean-drop something they don't understand, timed suspiciously with an IPO of a company.
11
52
1
510
11,398
Let’s also be clear that he created perfection spaces, and is currently executing a program to “patch” topology. Idk how laymen so confidently say things they have little expertise in.
Math is not just about proving results! He cut down in half the length of proofs for local Langlands, really simplified our understanding with Langlands-Kottwitz, p-adic geometry using perfectoid spaces. He is so good at mentorship that even his work with his postdocs (Stefanich with gestalten) and PhD students (Aoki with higher algebraic geometry and Wagner with physics/number theory) is revolutionary and all the buzz right now.
1
9
1,332
Perfectoid* - Ugh, auto correct.
1
136
Juspreet Singh Sandhu retweeted
Cameron--Martin theorem for SLE welding measures arXiv:2611.24875
1
1
250
I really tire with this adolescents that keep doing research math as if it is glorified-IMO. It really seems to me that in a field that praises Grothendick and Noether (among others) we are laced with the worst shadows of Erdos everywhere.
This guy is so stupid. It's unbelievable that he's the most public face of AI x Math. "We often think of mathematics as the highest prestige field ... the thing that smart people do to show they're smart" - no, YOU think of it as that because you've never grown out of your IMO phase. How come the highest prestige scientific field is the most underfunded if society thinks of it as that? "It's been hard for mathematicians to grapple with the fact that AI is better at them at math because it's a spiritual experience for most" - Yes, it is a spiritual experience for most (even society understands that), which means there's no measurable scale of being "better"— this is what we're arguing here. It's not a matter of grappling, the statement doesn't make sense. Keep up.
3
7
1
95
11,055
Juspreet Singh Sandhu retweeted
Recently at BIRS: Higher Rank Graphs in Algebra, Geometry and C*-Algebras birs.ca/events/2026/5-day-wo…
2
2
376
Juspreet Singh Sandhu retweeted
Tightness of bulk crossing heights for the planar discrete Gaussian free field Xiang Fang, Yi Xu arxiv.org/abs/2611.20051 [𝚖𝚊𝚝𝚑.𝙿𝚁]
1
2
126
Juspreet Singh Sandhu retweeted
Internal diameter bounds for DLA on random planar maps under invariance principles Deven Mithal arxiv.org/abs/2611.18540 [𝚖𝚊𝚝𝚑.𝙿𝚁 𝚖𝚊𝚝𝚑-𝚙𝚑]
1
2
67
Juspreet Singh Sandhu retweeted
As a grad student, it's very demoralizing to watch professors take cushy sabbaticals at AI labs, abandon advising duties, yet retain the prestige and decision-making power of their faculty titles. And the TCS community seems happy to turn a blind eye to this revolving door.
2
12
155
3,976
I concur with a lot of this - I’m happy @ketd47 has written this publicly.
This is true. Even more than a conflict of interest, I think it’s worth highlighting the hypocrisy of some of these people. When they would discourage their students from doing industry internships or before or even soon after ChatGPT first launched not applying for FREE significant GPU compute that their university was giving to faculty when the student repeatedly requested them to do so so that they could run LLM experiment ideas (which would eventually get published) saying “You’re giving up on academia which you shouldn’t do”. And then about turning two years later saying “I am about to become irrelevant by doing theory so I need to stay relevant by doing LLM work in AI labs” There’s also many others(not necessarily in the pictures) who are taking anti AI stances in public but secretly applying to AI labs trying to get jobs there. I don’t have a problem with this but be as honest in public as you are private otherwise you’re just a sneaky terrible person not fit to advise anyone.
6
510
Juspreet Singh Sandhu retweeted
Exactly! Our good shepherds have been used to create an illusion of peer-review, enough to generate headlines and keep the hype up, but with little more scientific value than a quick opinion. (Before the United Bots of AI come at my throat— I’m in the camp that thinks LLMs are amazing research tools, but science is not a videogame, and scientific norms are actually a pretty useful thing to adhere to. BTW want some good shepherding? Go public only when you understand what you have proved and have an employee who can give a talk and act as the corresponding author for a normal peer-review submission. Everyone will love you.)
3
10
1
71
4,451
Between Martin's "humanization" and @scottnarmstrong 's "informalization" we may be able to understand the scope & limitations of oAI's YM 4D work soon. Consider helping (since our seniors have graciously decided they must understand before us, lest our puny brains explode).
1
291
Juspreet Singh Sandhu retweeted
Stochastic Parallel Transport along Markov Processes: I. General Theory Kenta Misumi arxiv.org/abs/2611.11134 [𝚖𝚊𝚝𝚑.𝙿𝚁 𝚖𝚊𝚝𝚑.𝙳𝙶]
1
2
67
Juspreet Singh Sandhu retweeted
This is no exaggeration. Bałaban’s work on the UV stability of pure 4D Yang–Mills, including preparatory work, spans over 500 pages. If the new work constructs a gauge theory with fermions on all of ℝ⁴, even without a mass gap or confinement, it would be a major advance!
1,147 pages is amazingly short for a Balaban-style construction of four-dimensional gauge theory coupled to fermions. kitp.ucsb.edu/sites/default/…
3
5
40
6,936
Ok, when in South Korea last week, the foliage and food finally gave me the peace of mind to finish the first blog post on the PHAs 3/4 (in a series of 6) that I promised all of you. Here: juspreetsandhu.me/2026/08/05… 1/n
2
6
595
In this blog post, we take you on a "guided" tour (almost as it happened) on how the algorithm was designed, and the desiderata naturally emerged as we went from O(1) Wasserstein error -> O(1) TVD error -> o_n(1) TVD error. I also talk about 3 consequences of the work 3/n
1
113
I have started the second one (but we are also finishing off the WPI-SK + PHAs 2/5 to release soon, so speed is a little slow). The second blog post is on the key moment estimates proved by developing cavity theory, and observing cavity theory ≈ analytic subordination. 3/4
57