iiinova — where your digital self keeps evolving

Gautam Kamath @gautamkamath.com · 28/09/2026

Happy mid-autumn festival! 中秋节快乐! 🥮

Reposted by Gautam Kamath

Naomi Saphra @nsaphra.bsky.social · 24/09/2026

A pattern that I've been surprised by is people writing followup emails for their slop emails. Like, people are getting really insistent that I respond in earnest to emails they did not write.

4311

Reposted by Gautam Kamath

Aaron Roth @aaroth.bsky.social · 24/09/2026

AI tools are very useful for mathematical research. But at present, producing good work requires a labor intensive step that I have been calling "deslopping" - making the thing readable. Please do not skip this step; even if the theorem is great an unreadable paper is worthless.

4273

Gautam Kamath @gautamkamath.com · 24/09/2026

Anyone got a hot tip when #NeurIPS2026 decisions gonna drop 👀👀👀

110

Gautam Kamath @gautamkamath.com · 23/09/2026

NYU Courant professor Tristan Buckmaster, of Navier-Stokes drama fame, gave a talk at the new NYU Mathematics in the Age of AI seminar series. It was... popular.

012115

Gautam Kamath @gautamkamath.com · 22/09/2026

Over 100 years ago, steam engines were not welcomed by firefighters, who felt it would change what it meant to be a firefighter and it would result in job loss.

1100

Reposted by Gautam Kamath

Andrew Gordon Wilson @andrewgwils.bsky.social · 21/09/2026

I'm a Program Chair for ICML 2027. We'd love any creative suggestions for a great conference! How to manage the explosion of AI slop and insane submission growth? How to manage reviewing? How to incentivize high quality creative work? How to reduce bureaucracy and overhead?

10256

Reposted by Gautam Kamath

Transactions on Machine Learning Research @tmlrorg.bsky.social · 18/09/2026

TMLR now has a shiny new blog! By popular demand, the days of Medium are over. All the old posts are already ported, please check it out (and its beautiful colour scheme). Thanks to Managing Editor @paulvicol.bsky.social for setting it up! blog.tmlr.org

blog.tmlr.org

TMLR Blog

News, announcements, and editorial notes from the Transactions on Machine Learning Research (TMLR).

1112

Gautam Kamath @gautamkamath.com · 17/09/2026

Excited to join the new NYU Courant Institute School of Mathematics, Computing, and Data Science (@nyucourant.bsky.social) as an assistant professor of CS ~2 weeks ago! WOW its already been v exciting for ML and math at Courant in particular! While fun I hope not ever week is like this!!!

0460

Reposted by Gautam Kamath

Yoshitomo Matsubara @yoshitomo-matsubara.net · 16/09/2026

Do authors understand their own papers? Our desk rejection rate at TMLR is about 53% while it was only 6% in 2023 To understand the situation and assess our process, Nihar asked authors about their own TMLR submissions Learn the findings of the pilot study & share your thought

0234

Gautam Kamath @gautamkamath.com · 16/09/2026

Nihar Shah did a heroic experiment for TMLR: he spent 20-25 hours over two weeks interviewing authors of seemingly low-quality submissions about their own papers. He confirmed what we all suspected: people submitting these papers have *no idea* what is going on in them.

5276109

Reposted by Gautam Kamath

Transactions on Machine Learning Research @tmlrorg.bsky.social · 16/09/2026

TMLR has faced a deluge of submissions, necessitating stricter desk rejection policies due to limited reviewer capacity Co-EiC Nihar Shah reached out to authors of 10 papers slated for desk reject. Could they answer questions about their *own* submission? medium.com/@TmlrOrg/ask...

medium.com

Asking Authors About Their Own Papers

By Nihar B. Shah

116162

Gautam Kamath @gautamkamath.com · 15/09/2026

Two things are simultaneously true about math: 1. understanding is important 2. it's useful I could see a future where there are 2 parallel pillars of math (both important): one of human understanding, and one of incomprehensible Lean-slop used directly to do useful things

3183

Gautam Kamath @gautamkamath.com · 14/09/2026

Man folks really hate mathematicians huh. Seen some wild stuff on the Twitter timeline this past week. Including a 1M+ follower academic calling them "retards" and saying "Let AI wipe them out."

3211

Gautam Kamath @gautamkamath.com · 13/09/2026

In case you need a positive AI story right now: I have a baby monitor which automatically takes pictures and videos. Downloading is too slow: you have to manually click each one (infeasible for ~9k pics over 3 yrs). I had Codex write a script to download all of them in 45 mins.

1504

Gautam Kamath @gautamkamath.com · 11/09/2026

Even Bluey won't let you escape the AI safety debate

0160

Gautam Kamath @gautamkamath.com · 09/09/2026

When they tag @all or @channel in a Slack channel with over 1000 people 🚩🚩🚩

150

Reposted by Gautam Kamath

Conference on Secure and Trustworthy Machine Learning @satml.org · 09/09/2026

The submission server for #SaTML2027 is now open! There are also some changes to the Call for Papers, diffs are indicated on the website. Dates: Mandatory abstract deadline: Sep 22 Paper deadline: Sep 29 satml.org

195

Gautam Kamath @gautamkamath.com · 08/09/2026

Data privacy doesn't matter until it does

1190

Gautam Kamath @gautamkamath.com · 31/08/2026

As LLMs make things easier, we must raise our expectations on what researchers are expected to produce. This is particularly true for writing: clear writing was rarely a focus in scientific publication, and it's only got worse due to LLMs. We're trying to reverse this trend.

0314

Reposted by Gautam Kamath

Transactions on Machine Learning Research @tmlrorg.bsky.social · 31/08/2026

TMLR has refocused its acceptance criteria to place more emphasis on clear writing. Read more in this blog post: medium.com/@TmlrOrg/inc...

medium.com

Increasing Emphasis on Clear Writing in Acceptance Criteria

Three things are true today:

1357

Gautam Kamath @gautamkamath.com · 31/08/2026

Today's my last day at @uwcheritoncs.bsky.social, before I go on leave. I applied to 30+ faculty positions in Fall 2017. 3 interviewed me, only 1 offer. I'm deeply grateful that Waterloo CS alone saw potential here. I'm sad to be leaving and I will remain its loudest advocate.

3363

Gautam Kamath @gautamkamath.com · 26/08/2026

looking up the reference for Danskin's theorem

1131

Reposted by Gautam Kamath

Conference on Secure and Trustworthy Machine Learning @satml.org · 25/08/2026

The Call for Papers for #SaTML2027 is released! Deadlines: Abstract - Sep 22 Paper - Sep 29 Decisions - Dec 16 New this year: abstract deadline, initial review & ACs, mandatory artifact submission, planning for growth in submissions, etc. Full details: satml.org/call-for-pap...

122

Reposted by Gautam Kamath

Julian Togelius @togelius.bsky.social · 22/08/2026

A personal essay about how I’ve been feeling and thinking about this new technology that I’m contributing to and what it might to do to us all. togelius.blogspot.com/2026/08/losi...

togelius.blogspot.com

Losing my religion

In spring 2025 I had a crisis of faith. I thought about what the technology I'm helping to create might do to our future, and got scared. M...

99522

Reposted by Gautam Kamath

Anupam Gupta @anupamg.bsky.social · 22/08/2026

Another amazing result by Jason Li: an almost linear-time algorithm for maximum flows on capacitated directed graphs, avoiding the interior-point approach behind many of the recent max-flow results. arxiv.org/abs/2608.17384

arxiv.org

Maximum Flow Without the Outer IPM

We show that the balancing weights technique of Li (2026) actually produces an approximate *pseudo-circulation* of a directed, capacitated graph in $m^{1+o(1)}$ time. Together with standard flow techn...

2213

Reposted by Gautam Kamath

Conference on Secure and Trustworthy Machine Learning @satml.org · 21/08/2026

Reminder: #SaTML2027 has a Call for Competitions and Workshops! Both are due one week from today (August 28). Please submit your proposals and join us in beautiful Reykjavik! satml.org/call-for-com... satml.org/call-for-wor...

052

Gautam Kamath @gautamkamath.com · 20/08/2026

The Best Student Paper at COLT is named after Mark Fulk. I didn't know much about him, and information seemed scant online. Kaave Hosseini gathered information and wrote this short profile about him for University of Rochester CS. www.cs.rochester.edu/about/excell...

Mark Fulk died on May 18, 1997. Mark did his PhD work under Professor John Case at the University at Buffalo, graduating in 1985. Mark was a computer science faculty member at the University of Rochester from 1986 to 1993, after which he left to become a consultant at Genesee Valley Software company. He was the thesis advisor for two distinguished URCS PhD students, the late Laura Sanchez (who was a professor of computer science at Colgate University) and Sanjay Jain (who is a professor of computer science at the National University of Singapore).

Mark, an expert in inductive inference (recursion-theoretic learning theory), was praised at the department for his devotion to research, his enthusiasm, and his kindness to all. He was also a wonderful raconteur.

Mark is well known for several impressive results in inductive inference. Among his important contributions are:

Refuting Barzdin's conjecture that every robustly learnable class is recursively enumerable
Showing that prudence and reorder independence are not restrictive for explanatory learning
Showing the nonexistence of maximal identifiable classes
Introducing several interesting criteria of learning, such as inductive approximation, which is inspired by the scientific method, even if one may never reach a "perfect" model, at least reaching increasingly good models.
The best student paper award at the COLT (Conference on Learning Theory)—a conference that was held in Rochester in 1990 with him as the conference chair—is named in Mark's honor, reflecting his lasting influence on the field.

0110

Gautam Kamath @gautamkamath.com · 18/08/2026

Obviously this doesn't scale, but I chose to send this junior researcher a (hopefully polite) email with some of my thoughts (written hastily in 30 minutes, not polished). More practicing mathematicians & researchers should advocate for the culture they want, so here's my bit.

1234

Gautam Kamath @gautamkamath.com · 18/08/2026

I saw someone post a dozen+ AI slop papers purporting to solve niche open problems. This is antisocial behaviour and worse than if the problems stayed open. 0 people understand the solution, and there is no readable writeup. Also, incentive for either is removed.

2363

Reposted by Gautam Kamath

Foundations of Responsible Computing @forcconf.bsky.social · 17/08/2026

Recordings from FORC 2026 are now available! Please check them out. Also, subscribe to FORC's new YouTube channel while you're at it! www.youtube.com/playlist?lis...

youtube.com

FORC 2026 - YouTube

Talk recordings from FORC 2026

056

Gautam Kamath @gautamkamath.com · 14/08/2026

normalize putting something you already did on your to-do list just to check it off

2320

Gautam Kamath @gautamkamath.com · 13/08/2026

This was a tough decision. Surveys can have value. And I think a great survey now is of higher value than ever before. But finding great surveys is increasingly reviewer labour intensive, and it's adjacent to our main scope and mission.

3192

Reposted by Gautam Kamath

Transactions on Machine Learning Research @tmlrorg.bsky.social · 13/08/2026

Starting September 1, 2026, TMLR will no longer be considering survey papers for review. This is in part due to limited reviewer resources and significant required labour for reviewing surveys, as well as the evolving role of surveys in our field.

0126

Gautam Kamath @gautamkamath.com · 13/08/2026

I give my students advice on answering questions in talks, this viral Q&A illustrates what not to do. If you get a direct question, respond w direct answer, then elaborate. People remember the immediate response to their q the most. Don't build up to answer, def give an answer.

080

Reposted by Gautam Kamath

Clément Canonne @ccanonne.github.io · 11/08/2026

So... does the inequality holds for arbitrary (not just product) distributions with square Hellinger? ALMOST! 🤯 This is surprisingly non-trivial, and this is surprisingly true, and this is due to T.S. Jayram (2009): it's true, if you put a weird constant in front of the RHS! 5/

Chain Rule-Type Inequality for Squared Hellinger, similar to the KL one but with a constant 3.46 in front of the RHS.

1173

Reposted by Gautam Kamath

Sasho Nikolov @thesasho.bsky.social · 10/08/2026

Nikhil Bansal and I wrote a book-length survey of algorithmic and geometric methods in discrepancy theory, and a draft of it is online arxiv.org/abs/2608.00140. This book has been years in the making, and I’m excited to share it. More 👇

arxiv.org

Discrepancy Theory: An Algorithmic and Geometric Perspective

Combinatorial discrepancy theory is a subject with roots in combinatorics, geometry, and number theory, and with numerous applications to mathematics and computer science. At its core, discrepancy the...

1327

Gautam Kamath @gautamkamath.com · 10/08/2026

The significance & interestingness of most mathematical results can usually only be judged by a handful of experts in the area. Most people evaluating a result on social media can only do so via age of the problem and hype by the poster. Neither are great proxies for quality

1132

Gautam Kamath @gautamkamath.com · 10/08/2026

It was a pleasure to be one of the comms chairs for #ICML2026. The program chairs had an excellent vision throughout the process. I hope we were able to clearly communicate every decision and, whether you agree with them or not, why they were made.

040

Reposted by Gautam Kamath

ICML Conference @icmlconf.bsky.social · 10/08/2026

From behind the scenes: #ICML2026 comms chairs share what they were up to. Blogs, social media, interviews, press. First and foremost: a fundamental philosophy of transparency and openness. We hope you felt the same. Check out the post to learn more: blog.icml.cc/2026/08/10/w...

031

Gautam Kamath @gautamkamath.com · 10/08/2026

I'm pleased to share our #ICML2026 tutorial on machine unlearning! Presented by Vinith Suriyakumar and myself, it includes a full set of videos recorded and posted to YouTube! Please check it out: unlearning-tutorial.github.io

1187

Reposted by Gautam Kamath

Aaron Roth @aaroth.bsky.social · 07/08/2026

People used to be able to impress and intimidate reviewers with complicated proofs. This will change. In the age of AI inscrutable proofs are cheap. It is understandable proofs that are valuable. Opaque complexity is now it is a sign of laziness or lack of insight.

1638

Reposted by Gautam Kamath

Clément Canonne @ccanonne.github.io · 01/08/2026

"We're all worried," as what it means to do research (in my field, Theoretical CS) seems to be shifting, and shifting fast. What to do? Senior researchers must lead by example, knowing that not everything will pan out. What I'm suggesting below may not work everywhere, but here's my own advice: 1/

620351

Gautam Kamath @gautamkamath.com · 30/07/2026

ICLR 2027 has authorship quotas: - No more than 20 submissions per author - No more than 1 submission where no author has a paper previously accepted to a major ML conference From iclr.cc/Conferences/...

54813

Gautam Kamath @gautamkamath.com · 24/07/2026

We were the runner up for the Caspar Bowden PETS award. This recognized our work on differentially private fine tuning of language models. It was a big effort by a sizeable group across NA, Europe, and Asia. Thanks to the committee for the recognition.

2161

Gautam Kamath @gautamkamath.com · 24/07/2026

I've been referring people to this perspective every day for the last couple weeks. As we work together to determine new norms, we should not tolerate pure AI or otherwise bad writing. Tell your friends if they fall into this trap.

1172

Gautam Kamath @gautamkamath.com · 23/07/2026

Cheers from Calgary, where I am for #PETS2026! The general chairs have done a phenomenal job, you can see here they have "pets" (weighted helium balloons) you can adopt and take for a walk.

182

Reposted by Gautam Kamath

Conference on Secure and Trustworthy Machine Learning @satml.org · 23/07/2026

On top of our Call for Papers, #SaTML2027 has a Call for Competitions and a Call for Workshops (first time ever!!). Dates (easy to remember, same for both) Deadline: August 28, 2026 Notifications: September 18, 2026 Topics can be anything in trustworthy and secure ML!

112