iiinova — where your digital self keeps evolving

Thomas Steinke @stein.ke · 21/09/2026

It seems the free messaging-only inflight WiFi allows Bluesky (and not much else). Not sure if this is good news.

Thomas Steinke @stein.ke · 20/09/2026

In a US restaurant, you’ll get water without asking. In Germany, you have to ask twice before they finally relent and give you a glass of water. I bet the Brawndo gag was inspired by this.

010

Thomas Steinke @stein.ke · 11/09/2026

What’s the probability that I’ll gain weight over the next year? This question makes sense, but really it distracts from the fact that the outcome is mostly determined by my choices. “p(doom)” similarly distracts from our agency, although it’s a good way to get people’s attention.

2263

Thomas Steinke @stein.ke · 07/09/2026

AirBnB really needs to stop hosts from demanding personal details from guests. This didn't used to happen, but now half the time I'm being told I have to upload my driver license to some random website or my reservation will be cancelled.

180

Thomas Steinke @stein.ke · 06/09/2026

I took the kids out for a bike ride today & my bike fell apart - literally, a pedal fell off - so we were stranded. We walked to a restaurant for lunch and plotted how to get back home. The solution ended up being getting doordash to deliver a hex key to the restaurant & I managed to fix my bike. 😅

0130

Thomas Steinke @stein.ke · 28/08/2026

International convention for defining the proper functionality of Enter versus Shift+Enter

131

Reposted by Thomas Steinke

Zach Weinersmith @zachweinersmith.bsky.social · 23/08/2026

I was reading an author recently talking about how we don't appreciate the way that email isn't siloed. That is, you don't have to be in gmail to talk to someone else in gmail. The norm is so powerful, it'd be unthinkable to change it. All email is interoperable. Social media could've been this way.

4590999

Reposted by Thomas Steinke

Anthropic {bot} @anthropicai.xmirror.bot · 10/08/2026

We asked an unreleased research version of Claude to take a stab at the Riemann hypothesis. It didn’t solve it, but it did make strides on a related problem: it increased the lower bound for the fraction of zeros of the Riemann zeta function that satisfy the hypothesis from 41.6% to 67.2%.

anthropic.com

Learning more about Claude's mathematical capabilities

An unreleased version of Claude has made strides on a problem related to the Riemann hypothesis. It improved the lower bound for the fraction of zeros of the Riemann zeta function that satisfy the hypothesis, increasing it from 41.6% to 67.2%.

510513

Thomas Steinke @stein.ke · 05/08/2026

Woah! Even though I'm no longer at Google, this is shocking. (Obviously, ignore the heavy sugar coating in the official announcement.) blog.google/company-news...

blog.google

The next chapter of our AI momentum

Today, Google and Alphabet CEO Sundar Pichai shared some changes with Google DeepMind teams.

0120

Thomas Steinke @stein.ke · 04/08/2026

My radical suggestion for peer review: Move from nominal pre-publication review to explicit post-publication review. The original reason for peer review was that journal pages are a limited resource, so we need a filter before publication. That no longer makes sense with digital publishing. 1/

2215

Thomas Steinke @stein.ke · 25/07/2026

Fun fact: You can have two amazon accounts registered to the same email address, as long as they have different passwords. Don't ask how I discovered this or what happens if you change the passwords to be the same.

070

Thomas Steinke @stein.ke · 22/07/2026

xkcd.com/2385/ openai.com/index/huggin...

339150

Thomas Steinke @stein.ke · 19/07/2026

Buy new phone. Copy photos from old phone. Turn on cloud backup. "Your cloud storage is full." Delete photos from cloud. Photos deleted from phone because backup is on. 🫠 That is not what a "backup" is, Google.

1620311

Thomas Steinke @stein.ke · 18/07/2026

I couldn't find a nice (and up-to-date) visualization of the history of who is leading on arena.ai/leaderboard so I got Claude to make one: stein.ke/ai-leaderboa... (I'll try to keep this updated.)

stein.ke

Arena Leaderboard — Rating History

060

Thomas Steinke @stein.ke · 16/07/2026

What does “tokens per second per request” mean?? www.theatlantic.com/technology/2...

480

Thomas Steinke @stein.ke · 14/07/2026

Bastille Day 🇫🇷, Independence Day 🇺🇸, Canada Day 🇨🇦 are all in July. Isn’t it a nice coincidence that all these events happened in the middle of summer so we can enjoy national holidays in good weather. 😎 The pattern also holds down under: Australia Day 🇦🇺 is Jan 26, Waitangi Day 🇳🇿 is Feb 6.

030

Thomas Steinke @stein.ke · 02/07/2026

I’m attending COLT 2026 in San Diego. You can see the beach from the conference venue; not sure if that’s a good thing. 🤔 learningtheory.org/colt2026/

2170

Thomas Steinke @stein.ke · 01/07/2026

Today marks 10 years since I defended my PhD. Gosh, time flies. I can’t claim to be early career anymore.

070

Thomas Steinke @stein.ke · 29/06/2026

Why do you have two phones? Well, you see, sometimes there’s a QR code on a page and I need a second device to scan it.

2140

Thomas Steinke @stein.ke · 22/06/2026

If someone give you used baby stuff, the important thing is not whether it is useful to you, the important thing is that giving it away is a commitment device for them not having another baby.

050

Thomas Steinke @stein.ke · 12/06/2026

Do not make major life decisions without talking to people.

030

Reposted by Thomas Steinke

Gautam Kamath @gautamkamath.com · 04/06/2026

Honoured that our 2016 paper, Robust Estimators in High Dimensions without the Computational Intractability, w/ Ilias Diakonikolas, @daneelssoul.bsky.social, @jerryzli.bsky.social, Ankur Moitra, Alistair Stewart, was awarded the Gödel Prize This is the highest award for papers in theoretical CS 1/7

5486

Reposted by Thomas Steinke

Clément Canonne @ccanonne.github.io · 04/06/2026

Huge congratulations to Ilias Diakonikolas, Gautam Kamath, Daniel Kane, Jerry Li, Ankur Moitra, and Alistair Stewart on being awarded the Gödel prize for their breakthrough work on algorithmic robustness! www.sigact.org/prizes/g%C3%...

sigact.org

ACM SIGACT - Gödel Prize

1678

Reposted by Thomas Steinke

Differential Privacy Papers @dppapers.bsky.social · 28/05/2026

Privately Estimating Monotone Statistics in Polynomial Time Gavin Brown, Ephraim Linder, Mahbod Majid, Vikrant Singhal arxiv.org/abs/2605.27912

Privately Estimating Monotone Statistics in Polynomial Time

Gavin Brown, Ephraim Linder, Mahbod Majid, Vikrant Singhal

http://arxiv.org/abs/2605.27912

We study efficient differentially private algorithms for estimating monotone statistics, i.e., statistics that are monotone under the addition of new observations. The starting point for our investigation is subsample-and-aggregate: a classical paradigm that partitions the dataset into blocks, estimates the statistic on each block, and then privately aggregates the estimates.While practical and generically applicable, this approach is quite data-hungry. We improve upon this framework for the class of monotone statistics -- compared to subsample-and-aggregate, our algorithms save a factor of $t$ in sample complexity and pay a factor of $e^t$ in running time, where $t>0$ is a tunable parameter. We complement our results with a query-complexity lower bound, showing that our algorithms are essentially optimal for this task. As an application, we obtain improved results for private eigenvalue estimation, private loss estimation, and privately estimating a single parameter of a high-dimensional model, e.g., in linear regression.

041

Thomas Steinke @stein.ke · 27/05/2026

A handsome red bike, abandoned. By a busy road, outside my home. Locked to a tow away sign. What happened to its owner? Not even a thief has come for it. What will the HOA do? I see it daily and wonder.

A red bike in good condition locked to a signpost bearing a tow away sign, with a busy road in the background.

230

Reposted by Thomas Steinke

Jerry Chen @jcsalterego.bsky.social · 26/05/2026

but they were all of them deceived, for another monday was made

422211319

Thomas Steinke @stein.ke · 21/05/2026

You should only cite papers that you have read thoroughly. You should be confident that the papers you are citing are correct; you should also apply due diligence to the references therein. To be safe, you should probably not cite anything, unless maybe it's a self-citation.

2302

Thomas Steinke @stein.ke · 17/05/2026

I can't believe that in 2026 bibliographies are still more likely to tell you which page of the printed journal the article is on or which city the publisher is based in than to give you a URL. Were page numbers *ever* useful? Did print journals not have a table of contents?

4242

Reposted by Thomas Steinke

Thomas Steinke @stein.ke · 16/05/2026

Here's an example of a "hallucinated" citation from a decade ago (i.e. pre-LLMs). The same bad citation appeared in multiple papers. I eventually traced the source to Google Scholar (and it's now fixed).

[2] Marcus Hardt and Jonathan Ullman. Preventing false discovery in interactive data analysis is
hard. In Foundations of Computer Science (FOCS), 2014 IEEE 55th Annual Symposium on, pages
454–463. IEEE, 2014.

From: https://arxiv.org/abs/1510.03349

[11] Marcus Hardt and Jonathan Ullman. Preventing
false discovery in interactive data analysis is hard.
In Foundations of Computer Science (FOCS), 2014
IEEE 55th Annual Symposium on, pages 454–463.
IEEE, 2014.

From: https://proceedings.mlr.press/v51/russo16.pdf

[19]
Marcus Hardt and Jonathan Ullman. 2014. Preventing false discovery in interactive data analysis is hard. In IEEE Symposium on Foundations of Computer Science (FOCS). 454--463.

From: https://dl.acm.org/doi/10.1145/3139550.3139556

Article (Correct)

Preventing False Discovery in Interactive Data Analysis Is Hard
Authors: Moritz Hardt, Jonathan Ullman
FOCS '14: Proceedings of the 2014 IEEE 55th Annual Symposium on Foundations of Computer Science
Pages 454 - 463
https://doi.org/10.1109/FOCS.2014.55
Published: 18 October 2014 Publication History

From: https://dl.acm.org/doi/10.1109/FOCS.2014.55

272

Thomas Steinke @stein.ke · 16/05/2026

While I generally agree with this policy, the focus on hallucinated citations specifically makes me uncomfortable: I usually get my BibTeX from Google Scholar. Yes, GS is sometimes wrong. No, I don't carefully check my references. Is using AI to fill in references really morally any different?

791

Reposted by Thomas Steinke

Mark Riedl @markriedl.bsky.social · 14/05/2026

ArXiV has a new LLM policy (Screenshots with alt text so you don’t have to click through to the other place and see all the stupid responses)

Attention @arxiv authors: Our Code of Conduct states that by signing your name as an author of a paper, each author takes full responsibility for all its contents, irrespective of how the contents were generated. 1/

If generative AI tools generate inappropriate language, plagiarized content, biased content, errors, mistakes, incorrect references, or misleading content, and that output is included in scientific works, it is the responsibility of the author(s). 2/

We have recently clarified our penalties for this. If a submission contains incontrovertible evidence that the authors did not check the results of LLM generation, this means we can't trust anything in the paper. 3/

The penalty is a 1-year ban from arXiv followed by the requirement that subsequent arXiv submissions must first be accepted at a reputable peer-reviewed venue. 4/

Examples of incontrovertible evidence: hallucinated references, meta-comments from the LLM ("here is a 200 word summary; would you like me to make any changes?"; "the data in this table is illustrative, fill it in with the real numbers from your experiments") end/

8312105

Reposted by Thomas Steinke

Clément Canonne @ccanonne.github.io · 11/05/2026

List of accepted papers at #COLT2026, the Annual Conference on Learning Theory: learningtheory.org/colt2026/acc... (h/t @gautamkamath.com)

learningtheory.org

COLT 2026 - Accepted Papers

0115

Thomas Steinke @stein.ke · 10/05/2026

I did a back-of-the-envelope calculation to satisfy my curiosity: A standard gasoline/petrol pump delivers about 15 megawatts of energy. (In contrast, the fastest electric vehicle chargers deliver half a megawatt.)

020

Thomas Steinke @stein.ke · 10/05/2026

My cynical hypothesis for why there's a lot of discussion about datacenters in space is that Elon needed a justification for merging xAI with SpaceX. Elon being the CEO of both is an obvious conflict of interest, so he needed to explain to investors why it made sense -- hence space datacenters.

1130

Thomas Steinke @stein.ke · 05/05/2026

Pleased to share that I'll be presenting this paper at COLT in San Diego! 😁

150

Thomas Steinke @stein.ke · 05/05/2026

Re. AI "consciousness" takes: I would be interested to read something from a psychiatry or psychology perspective. People who have spent time debugging human intelligence might have something insightful to say about artificial intelligence.

240

Thomas Steinke @stein.ke · 04/05/2026

I remember watching some American lecture a New Zealand fast food worker about this view. As far as I can tell this is an Americanism and totally arbitrary.

130

Thomas Steinke @stein.ke · 01/05/2026

I have an app to control my garage door opener. Every time it takes like 30s to load. What is it loading? Once I'm logged in, it should only need to send one UDP packet: (app_id,door_id,open_or_close,timestamp,hash(app_secret_key,door_id,open_or_close,timestamp)) That's like 32 bytes tops.

000

Reposted by Thomas Steinke

Clément Canonne @ccanonne.github.io · 28/04/2026

Nothing incredible novel, but a short 🧵 thread about MGFs, their use, and why I love Taylor series. Say you want to compute the moments of a r.v. X, which is "sufficiently well-behaved" (in some rigorous sense). That is, 𝔼[Xⁿ], for some integer(s) n≥0 of your choosing. How do you proceed?

3277

Thomas Steinke @stein.ke · 17/04/2026

AC: Please discuss this submission and respond to the authors' rebuttal. me: The scores range from "reject" to "strong reject". We don't need to waste more time on this.

060

Reposted by Thomas Steinke

Peter Henderson @peterhenderson.bsky.social · 27/03/2026

This is a challenging legal problem for NeurIPS (and other conference participants)! You might be wondering how this is possible given the First Amendment? I wrote a quick explainer on the current status quo of relevant First Amendment cases & law to get you up to speed. 🔗👇

185

Thomas Steinke @stein.ke · 25/03/2026

My parents are the opposite: They check in, drop bags, & then go wait at home. "Boarding is in 10min." "OK, let's eat dessert." "No, we need to drive to the airport now!" "Relax, they have your bags; they won't leave without you." *30 min later* "Final call for passenger Steinke." Stressmaxxing...

140

Thomas Steinke @stein.ke · 17/03/2026

ICML requires authors to serve as reviewers. Reviewers aren't supposed to use LLMs (with caveats 🧵). ICML used prompt injection to catch reviewers using LLMs. And now they have desk rejected the papers submitted by those reviewers. What a mess...

icml.cc

ICML 2026 Call for Papers

4275

Thomas Steinke @stein.ke · 17/03/2026

Schools have ruined St Patrick's and Valentine's days for parents. E.g., this morning I had to prepare lunch for a random other kid in my kid's class. What does this have to do with St. Patrick?

110

Thomas Steinke @stein.ke · 15/03/2026

Which task should I do first? (i) Send reminder emails to late reviewers. (ii) Submit my own late reviews.

1101

Thomas Steinke @stein.ke · 12/03/2026

Some people are still debating whether or not LLMs are "useful," so let's stake out one clear use case: LLMs are useful for translating between languages. That includes translating between natural languages (e.g. Spanish to English) and, more recently, to formal languages (e.g. English to Python).

Screenshot of bsky post with author's identity not shown. [I'm not trying to pile on.]

Sure, LLMs are useful for:
1. Fraud
2. Plagiarism
3. Cognitive off-loading
Which of those use-cases are you promoting?

3:59pm March 10 2026
51 reposts, 14 quotes, 257 likes, 6 saves

3332