I just wish I could make the equivalency network of ideas and tools and missions real and interpersonal and convivial

i submitted a revision last night and bought a graphic fleece in celebration and the shopkeep had to point out that it looks just like the one i was wearing at the time fml me: Leave me alone in my pile of cozy extremely graphic fleeces

Academic fields differ in their degree of consensus about what "better" publication venues are — here shown both for consensus about the "top 5" and for pairwise choices. Hello #econsky :) arxiv.org/abs/2603.00807 @jugander.bsky.social @danlarremore.bsky.social @aaronclauset.bsky.social

I often write "sermons" to our applied math PhD students as part of my Friday roundup email. This week, I wrote about the use of AI (and specifically LLMs) in applied math research: www.cs.cornell.edu/~bindel/serm...

feeling like an interdisciplinary leper living somewhere in the black hole of "you're more of a x than a y" (says y) and a "you're more of a y than an x" (says x) ☠️

Benchmarks Saturate When The Model Gets Smarter Than The Judge arxiv.org/pdf/2601.19532 may be worth a look? Recent preprint which looks specifically at LLM-as-judge failures for the Omni-MATH dataset

We actually found (following some earlier work) that the pairwise/pointwise difference does not actually matter that much, and the intuition that pairwise judgments are better because they have a reference point may not hold true aclanthology.org/2025.finding... aclanthology.org/2025.emnlp-m...

gsood.com/research/pap... This may be helpful on both fronts! + refs in the lit review

Dear hivemind, Do you have any favorite references on failures of LLM-as-a-judge (and universally enforceable validity checks? besides some kind of cohen's kappa on a small subset)

In this new report, @ksavin.bsky.social, and @calliefreitag.bsky.social and I explain how Trump 2.0 and DOGE built up new barriers to the already-difficult process of getting disability benefits. Let's talk about what we found, and what we recommend to address the issues. 🧵

they added an AI admin to review my concur request 😁

Difficulties accessing disability benefits: “Getting stuck in a loop.” We identified five major areas... (a) changes to the phone system, (b) field offices’ appointment and walk-in policies, (c) staffing cuts and reassignments, (d) increased processing times and denials, ..

It's been a journey of nearly 3 years, but I'm very excited to announce the CNTR AISLE Portal! 🚀 cntr-aisle.org It’s a new way to review and evaluate the 1,000+ AI bills introduced in the U.S. over the last three years. Check out the Bill Library and our Profiles#AIPolicy #OpenData

www.nytimes.com/2026/02/25/n... Big news - allegheny county's erin dalton tapped for ny social services chief role

Today arXiv remembers our colleague Joe Halpern, who was instrumental in founding arXiv's CS section. Joe's passions ranged far & wide and we're lucky that arXiv was one of them. Joe, thank you for giving so much to arXiv - you are missed. blog.arxiv.org/2026/02/27/remembering-joe-halpern

You can kill the product but not the idea - today, 160 Congressional co-sponsors introduced a bill to bring back the IRS’ much beloved free tax filing product, Direct File. www.theverge.com/policy/88522...

Acemoglu, Autor & Johnson sketching out a pro-worker AI agenda. Worth your time.

catching up on superbowl ads. www.youtube.com/watch?v=sCyi... the nerds are out of touch 🫣 just show me ai that will do my laundry 🫣

Took a break from my 25+ pg and growing response to referees to yak shave a rebuttal down to 5000 chars 🐑✂️🙏💆‍♀️

When I’m two weeks behind on my grading.

wish there could be like the equivalent of knitting or crafting circles where reasonable folks who are reasonably skeptical and tired of ai hype can get together and build little better futures together

A key challenge, and a huge source of the disconnect between Silicon Valley and the other creative industries is that, for coders, AI eliminates the drudgery so they can focus on the creative/expressive part. In every other creative discipline, it has the exact opposite effect.

It’s worth reading the op-ed Zoë Hitzig wrote about the AI advertising landscape. Very thoughtful and informative

If this kind of work speaks to you, come work with us! My team at CAISI is hiring an Applied Systems AI Research Scientist, among many other roles. www.nist.gov/caisi/career...

Last chance to present a paper at this Econ-CS conference #econsky marketdesigner.blogspot.com/2026/02/dead...

it is part of an ongoing crowdsourced data project if you would like to contribute: docs.google.com/forms/d/e/1F...

urbanomnibus.net/2026/02/wher... This is a very cool project reverse engineering Amazon shipping routes!

👀💆 my favorite genre: mutual aid / cooperativism operational/logistic efficiencies

This Equifax takedown letter from Senators Warren, Wyden, and Sanders is encouraging, but it's insane we need Senators to demand public price transparency from a contractor that works for the public

github.com/tobi/qmd this helps a bit with token guzzling if you're having it read a bunch (you can always fall back on ripgrep if embedding search fails)

re: token guzzling... (you may already know this, but in case it helps anyone else reading this)... the default "heartbeat" for OpenClaw is 30 minutes, which is quite token-heavy, and people on [other place] are reporting that setting it to more like 4 hours is a good task-pace-to-token tradeoff.

If you use letta.bot instead through our API, we bill per step, not per token. GLM 4.7 and MiniMax M2 are all free, and the $20 a month plan is model agnostic and has a ton of usage.

some observations after a few days with my molty/clawdbot (called magikarpy) on an old laptop --- a telegram/messaging interface to a filesystem + claude code is genuinely useful --- maybe not for core work where you need to know what's going on, but great for side projects --- guzzles tokens

...........,..,.........im dying/dead......its over

SCOOP: Palantir's AI has been deployed at HHS since last March to audit grants, grant applications, & job descriptions for anything related to "DEI" or "gender ideology" The goal was to ensure compliance with 2 of Trump's executive orders: www.wired.com/story/hhs-is...

An LAPD officer stands on the corner of Temple and Alameda facing Barbara Kruger's iconic mural featuring 9 questions: “WHO IS BEYOND THE LAW? WHO IS BOUGHT AND SOLD? WHO IS FREE TO CHOOSE? WHO DOES TIME? WHO FOLLOWS ORDERS? WHO SALUTES LONGEST? WHO PRAYS LOUDEST? WHO DIES FIRST? WHO LAUGHS LAST?”

i splurged on an eink monitor to crank through my backlog of projects to ship out the door. it's great. every day just trying to get closer to a liberation from screens

If you're reading about Epstein and his friends today, could you also take time to read this document? I guarantee that someone at your university right now is a sexual harasser and/or abuser. Maybe you could stop them.

what's really sad is that the AI apocalypse is shaped like reddit

This is not true; I beg people read the full paper and especially the study design. The conclusions mirror my (and many other practitioners') conclusions: if you use AI critically and engage both with the question and the answer, it has a net positive impact on both learning and productivity

The latest from Anthropic: using Anthropic's products makes you worse at your job

I wrote about Clawdbot/Moltbot/OpenClaw and Moltbook, the fascinating, weird and sometimes even useful social network for digital assistants to swap tips and gossip with each other simonwillison.net/2026/Jan/30/...

Microsoft Research NYC is hiring a researcher in the space of AI and society!

CS ArXiv recently banned “review and position” papers, but what are those? Do they include more generated content? Who is most affected by this change? @yanai.bsky.social and I dug into the data to find out! Nearly 50% of Computers & Society papers might be censored, vs 3% of Computer Vision ‼️

Hello! This is a reminder that @cornelltech.bsky.social runs a Red Team Clinic that provides a *free* safety consultation to nonprofits / public sector orgs that are developing a public-facing AI tool and want to stress-test it for possible abuse vectors. Applications welcome on a rolling basis:

In the piece I argue that we can Enforce, Build, and Simplify our way out of this Means-Testing Industrial Complex and stop sacrificing our public programs to poorly performing contractors.

The story is the same for Deloitte's domination of eligibility systems that determine who gets Medicaid coverage and who does not. With over $6B in contracts across 25 states, new work requirements included in OBBBA are an enormous opportunity to prey on bureaucratic complexity and profit.

This Means-Testing Industrial Complex is epitomized by Equifax's domination of income verification where the government has backed itself into a corner. OBBBA requires more complex and frequent verifications which will rely on one dominate private company that is mercilessly hiking prices.

Corporations are extracting billions from America’s safety net and their payday is about to be supercharged by new rules. I lay out how programs like Medicaid have become vulnerable to corporate capture and what we need to build to stop it. @lpeblog.bsky.social lpeproject.org/blog/the-mea...