Sign in

Nikhil Garg

@nkgarg.bsky.social
3K followers 1.4K following 463 posts

I study algorithms/learning/data applied to democracy/markets/society. Asst. professor at Cornell Tech. gargnikhil.com. Helping building personalized Bluesky research feed: bsky.app/profile/paper-feed.bsky.so…

PostsRepliesMedia
Nikhil Garg @nkgarg.bsky.social · 14/10/2026
wonderful, congrats!
010
Nikhil Garg @nkgarg.bsky.social · 14/10/2026
(In our updated version to be uploaded soon, we do apply it to a bunch of abstracts from 2026, and could easily apply it to all....though we're not going to publicly share individual-level results.)
110
Nikhil Garg @nkgarg.bsky.social · 14/10/2026
if the goal is just arxiv abstracts, then building a classifier that performs ~as well as Pangram is easy. (their big impressive thing is bringing a lot of domains in-distribution). We do that in this paper, though I would stay away from individual labels... arxiv.org/abs/2606.25152
arxiv.org
Hitting a Moving Target: Test-Time Adaptation for AI Text Detection under Continual Distribution Shift
Deployed approaches for AI text detection often rely on training-time access to labeled datasets of both human-written and AI-generated text. This approach is vulnerable to three types of distribution...
110
Nikhil Garg @nkgarg.bsky.social · 13/10/2026
These two passages were super interesting, and getting at what I think is these models' superpower -- tenacity to try everything. (Alongside all the right open questions regarding credit)
Are we the only ones that were close to this result? I think not. I think that other people might have had the same idea as us and could have been 1 or 2 Claude-hours away. Or, if they had more time, they could have pushed the idea through more tensors (instead of abandoning the idea on overleaf for years) and would have inevitably tried Coppersmith's paper, and we would have gotten a human proof. But none of us pushed hard enough and a machine did. 

In any case, who gets credit anymore? Is it the machines who assemble all of our joint knowledge and ideas and plow through them with unnerving speed (and $$cost$$)? Is it the people who prompt the machines (i.e., "push the button")? Does a prompt such as "go solve the problem" without any extra hint count? Is it the people who communicate the new knowledge most effectively, and what does that mean? Who gets to publish the new work? Do publications mean more than just exposition? [I am actually pretty reluctant to submit our write-up to STOC, even though I know that most papers will have AI generated results. Our community is so divided about credit and publications that I have started to question the need for STOC as a publication venue, but I digress.]
060
Reposted by Nikhil Garg
Clément Canonne @ccanonne.github.io · 13/10/2026
Virginia Vassilevska Williams wrote some thoughts about the 3-SUM breakthrough Josh Alman and her wrote an exposition of, how it happened, what it means, and some context about 'what could have been' in a world with more time and fewer models. Read it here: people.csail.mit.edu/virgi/though...
people.csail.mit.edu
https://people.csail.mit.edu/virgi/thoughts.txt
A week has passed since Josh and I posted our write-up on the arXiv, and I decided to write a few things. Much was said about our write-up. Some were excited, others enraged. Most everyone was scar...
1396
Nikhil Garg @nkgarg.bsky.social · 11/10/2026
Amazing, congrats and good luck!
020
Reposted by Nikhil Garg
Graze Social @graze.social · 07/10/2026
🗳 midterms.graze.social is @graze.social's most ambitious project to date. And as of today, it's live. Let's take a look at what the next 27 days will look like 🧵
midterms.graze.social
Election Night 2026
Vetted coverage of the 2026 midterms, by state.
513851
Nikhil Garg @nkgarg.bsky.social · 06/10/2026
Very excited to have this out -- @kennypeng.bsky.social 's brainchild in thinking about the implications of sparse representations. He's on the job market, and watch out for some exciting tools we're releasing soon! arxiv.org/abs/2610.05794
0111
Reposted by Nikhil Garg
Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 30/09/2026
In our Nature paper, we introduce the first superhuman Stratego AI, which we built using general techniques that we developed for RL & test-time compute under imperfect information. www.nature.com/articles/s41...
nature.com
Scalable decision-making for games of imperfect information - Nature
Ataraxos, an AI for the board wargame Stratego, establishes a design pattern for reinforcement learning and search that is effective under large amounts of hidden information, a longstanding desiderat...
1225960
Nikhil Garg @nkgarg.bsky.social · 30/09/2026
a fun fact is that several (at least 3?) of the authors were/are Jon Kleinberg's students at the same time at Cornell
130
Reposted by Nikhil Garg
Isabel Silva Corpus @isabelcorpus.bsky.social · 30/09/2026
In 2023, Change.org integrated an AI-assisted writing tool into their platform. We found that access to the tool changed platform trends in petition text, but that outcomes (signatures and comments) did not improve... Excited that this paper is out, check it out! 🔗: www.nature.com/articles/s41...
nature.com
Introducing AI to an online petition platform changed outputs but not outcomes - Nature Human Behaviour
Corpus et al. examine how the introduction of an AI writing tool impacted Change.org petitions, finding that it increased petition homogeneity but did not improve petition outcomes.
22412
Reposted by Nikhil Garg
Clément Canonne @ccanonne.github.io · 29/09/2026
The report from the AI + TCS Working Group held at the @simonsinstitute.bsky.social is out, including "twelve concrete near-term recommended actions." Currently reading it: while I have some reservations, but there is some good in it! Notably items 7, 8, and 11. simons.berkeley.edu/ai-tcs-worki...
7. Best works and technical talks in interviews.

- Emphasize quality over quantity. Evaluate candidates on no more than three  works of their choosing, with this fact explicit in the instructions for candidates and their letter writers.

- Test technical depth. Evaluate the depth of the candidates’ understanding of the aforementioned works more than was done normally until now.

8. Additional postdocs. Establish additional postdoctoral positions for the TCS community, a temporary measure that will be phased out after a few years.

9. Workshops for identifying new theory directions. First, organize a workshop (or several) with the goal of identifying a list of modern topics related to AI; and second, run a series of workshops on each of these topics to identify fundamental directions and establish foundations and agendas for future work.

10. “TCS in the Age of AI” series at the Simons Institute. Establish a recurring and widely accessible “TCS in the Age of AI” series at the Simons Institute.

11. Opportunities for training and evaluation without AI assistance. To ensure development of necessary skills for a successful career in academia, a PhD in TCS should partially include training and evaluation components where the use of AI is forbidden.
43112
Reposted by Nikhil Garg
Maria Antoniak @mariaa.bsky.social · 28/09/2026
behold! #colm2026 with a new filter 👀 on lea.ac
A screenshot of Lea showing the COLM conference page with the "Papers" tab selected. The listed papers are filtered to those with authors whom I follow on Bluesky / ATProto.
0184
Reposted by Nikhil Garg
Daniel de Kadt @dandekadt.bsky.social · 24/09/2026
Very excited: Cornell Arts & Sciences seeks up to 5 postdoctoral scholars working on AI and the humanities/social sciences. Not involved in selection but could serve as a faculty sponsor if you are working on topics related to ~"how to do social science in the age of AI." Happy to chat.
academicjobsonline.org
Cornell University, College of Arts and Sciences Fellowships
Job #AJO32791, Postdoctoral Associate, AI in Social Sciences & Humanities, College of Arts and Sciences Fellowships, Cornell University, Ithaca, New York, US
11715
Reposted by Nikhil Garg
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.
5293117
Nikhil Garg @nkgarg.bsky.social · 15/09/2026
Not the most important question here, but: do you think these papers should be accepted to, say, STOC/EC and given the same treatment that the equivalent paper 1 year ago would have received?
120
Nikhil Garg @nkgarg.bsky.social · 15/09/2026
agreed. I'm not sure where this fits in to your list: in the "applied applied math" areas I work in, the understanding is what is useful -- the hard-core general theorems aren't important, but having thought through the models carefully is important for market/algorithm design intuition
030
Reposted by Nikhil Garg
Aaron Roth @aaroth.bsky.social · 15/09/2026
A world without open problems Here are some that fell today: K-server: arxiv.org/abs/2609.15979 Matroid Secretary: arxiv.org/abs/2609.145... Matrix Spencer: arxiv.org/abs/2609.15025 (Well Matrix Spencer was maybe also a few weeks ago, but who's counting? arxiv.org/abs/2608.28816 )
arxiv.org
The $k$-server conjecture is true
The $k$-server conjecture states that a deterministic online algorithm can achieve competitive ratio $k$ on every metric space. We prove the conjecture. Specifically, we show that the work function al...
3316
Reposted by Nikhil Garg
Tony Zhou @tyzhou.bsky.social · 14/09/2026
Bluesky gives users the power to choose their algorithms, but how does this actually work in practice? In our new preprint, we explore the ecosystem of Bluesky custom feeds and the feed creators behind them! 🧵 arxiv.org/abs/2609.12958
arxiv.org
Middleware for Feed Recommendation in Practice: How Feed Creators Build, Maintain, and Sustain Custom Feeds on Bluesky
Scholars have long proposed third-party middleware as an alternative to centralized algorithmic feeds: feeds built and distributed by independent feed creators. This vision saw no large-scale instanti...
513546
Nikhil Garg @nkgarg.bsky.social · 12/09/2026
Very cool and important work! Curious to see results
032
Nikhil Garg @nkgarg.bsky.social · 02/09/2026
Most data colada posts combine two basic insights. (a) these numbers contradict (e.g., average scores aren't averages of their constituent parts). (b) these numbers are inconsistent with reasonable data generating processes (effect sizes too large; humans like round numbers).
020
Nikhil Garg @nkgarg.bsky.social · 02/09/2026
(2) A good (non-erroneous) analysis will be self-consistent -- different views of the same data should tell the same story. My most reliable bug catching technique is "these 2 numbers are inconsistent. I don't know which, but one of these must be wrong". and this is why fraud is hard!
110
Nikhil Garg @nkgarg.bsky.social · 02/09/2026
I love reading data colada posts. Besides their importance in catching fraud, many implicit lessons for what good data science looks like. (1) curiosity + simple methods go very far. Most of their posts are "we took this average and stared at the raw data, and funny things popped out".
1171
Nikhil Garg @nkgarg.bsky.social · 31/08/2026
This all rings true to me. I still credit my summer of Kumon (in early 2000s) as accelerating my math education by several years. (and then my parents wanted to stop paying for it)
051
Nikhil Garg @nkgarg.bsky.social · 29/08/2026
It's the only one I read but I thought it was quite good. Before it I was only familiar with his computer architecture and game theory contributions
140
Nikhil Garg @nkgarg.bsky.social · 29/08/2026
Have you read The Man from the Future? I thought it was pretty good but maybe you want something more exhaustive? www.goodreads.com/book/show/61...
goodreads.com
The Man from the Future: The Visionary Ideas of John vo…
The smartphones in our pockets and computers like brain…
130
Nikhil Garg @nkgarg.bsky.social · 28/08/2026
I'm curious how you think we should share errata, of others work or our own. Clearly, we're about to find way more of these than we ever have before. Retractions seem extreme for most cases, and official errata may drown editorial processes
120
Nikhil Garg @nkgarg.bsky.social · 28/08/2026
Yeah, agreed, I'm almost through auto formalizing most of my work in lean and working through the issues. It's scary but imo we all need to figure out an errata standard!
050
Reposted by Nikhil Garg
Wesley Finck @wesleyfinck.org · 26/08/2026
Some preliminary product insights from our recent onboarding launch: It's no coincidence that the most popular suggested link is "From Feeds to Trails". People are ready for a new kind of social media experience! leaflet.pub/p/did:plc:ld...
leaflet.pub
From Feeds to Trails
To design the future of social media, rethink the interface before the algorithm.
3255
Nikhil Garg @nkgarg.bsky.social · 26/08/2026
Wonderful, congrats!!
020
Reposted by Nikhil Garg
Nikhil Garg @nkgarg.bsky.social · 19/08/2026
Then, we worked with NYCPS to nudge applicants to apply to high-performing schools, while being congestion aware! The paper (full version out soon) won Best Paper with Student Lead Author at EC`26 -- see Erica's great thread here! bsky.app/profile/eric...
141
Nikhil Garg @nkgarg.bsky.social · 19/08/2026
Thanks for Grace at the Cornell Chronicle for writing about our work, and to our wonderful partners at NYCPS! This project has been one of the most stressful but also rewarding ones of my career, and I think a testament to the value of slow, application-driven work news.cornell.edu/stories/2026...
news.cornell.edu
Research helps NYC students aim higher in public high school applications | Cornell Chronicle
Many New York City eighth-graders – particularly those from underserved communities – aren’t applying to academically competitive high schools where they could succeed, according to new Cornell resear...
020
Nikhil Garg @nkgarg.bsky.social · 19/08/2026
Then, we worked with NYCPS to nudge applicants to apply to high-performing schools, while being congestion aware! The paper (full version out soon) won Best Paper with Student Lead Author at EC`26 -- see Erica's great thread here! bsky.app/profile/eric...
141
Nikhil Garg @nkgarg.bsky.social · 19/08/2026
Would love your thoughts!
010
Nikhil Garg @nkgarg.bsky.social · 18/08/2026
Thank you, would love comments :)
010
Nikhil Garg @nkgarg.bsky.social · 18/08/2026
New paper out, for a project 3+ years in the making! First, in this paper, we quantified sources of disparities in NYC high the NYC high school match -- see Kenny's great thread here! www.nature.com/articles/s44...
nature.com
Connecting application behavior to undermatching in New York City school choice - Nature Cities
Cities are challenged to allocate educational resources efficiently and equitably. This study of the 2022–2023 New York City High School Match program finds that Black and Hispanic students undermatch...
2303
Reposted by Nikhil Garg
Jason Koebler @jasonkoebler.bsky.social · 18/08/2026
“X’s algorithm learns what you hate and shows you more of it, according to a new study just published in the Proceedings of the National Academy of Sciences” www.404media.co/xs-algorithm...
404media.co
X's Algorithm Feeds Off Ragebait and Impacts Democrats More, Study Finds
X is driving engagement by making users fight in the replies.
1042139902
Nikhil Garg @nkgarg.bsky.social · 18/08/2026
Congrats and good luck!
020
Nikhil Garg @nkgarg.bsky.social · 18/08/2026
Same -- I'm doing in class quizzes and an oral exam to explain project code. I say they're not allowed to use it for writing their project report, but instead of policing it I will just warn them that AI is not good at what we're looking for (concision, clarity, reflecting course content)
020
Nikhil Garg @nkgarg.bsky.social · 17/08/2026
Ah yes should have assumed you were aware :)
110
Nikhil Garg @nkgarg.bsky.social · 17/08/2026
I'm not sure who all is actually in the room across conferences, but Nihar Shah (www.cs.cmu.edu/~nihars/) has published a bunch + helped run experiments + set policy on conference peer review, including recently on submission limits per author and reviewer assignment to prevent collusion
cs.cmu.edu
Nihar B. Shah - CMU
Nihar B. Shah, Associate Professor in MLD and CSD at CMU.
140
Nikhil Garg @nkgarg.bsky.social · 15/08/2026
yeah, I broadly feel that way (more excited) but 1) have always felt that way about the role of theory/applied modeling; and 2) am reasonably established in my career
110
Nikhil Garg @nkgarg.bsky.social · 15/08/2026
I absolutely understand it among theorists (it's a real shock to what skills may be valued, and how one measures progress) but tend to agree for more building-oriented fields.
130
Nikhil Garg @nkgarg.bsky.social · 14/08/2026
I've logged in and giving it a serious go! Crazy impressive work by them + cool that the protocol supports innovation like this
1244
Nikhil Garg @nkgarg.bsky.social · 14/08/2026
I know Sophie is traveling, but I'll try to make it!
020
Nikhil Garg @nkgarg.bsky.social · 14/08/2026
can't wait to try this out in Atlanta in December!
020
Reposted by Nikhil Garg
Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 12/08/2026
For the past six months @mariaa.bsky.social and I have been working on creating an improved research experience on atproto. It's now live! Check out: lea.ac
leaflet.pub
Lea: research-oriented social media
Introducing a beta-version of our new research-focused social media platform
20381113
Nikhil Garg @nkgarg.bsky.social · 02/08/2026
Very interesting thread on, among other things, human verification and understanding of the latest AI proofs
070
Nikhil Garg @nkgarg.bsky.social · 01/08/2026
I agree with that snippet, and it implies a large change in norms or processes -- will papers be published, and who will author them?
010
Reposted by Nikhil Garg
Jonathan Stray @jonathanstray.com · 31/07/2026
Would you like to try the first ever algorithmic feed with CONTROLS? GreenEarth is built to serve you, not the platform or anybody else. Now in beta. www.greenearth.social/p/welcome-to...
greenearth.social
Welcome to the GreenEarth Feeds
Social media algorithms built for people who hate social media algorithms
1134