Jeremy Gibbons

Just in case this Norwegian Blue shuffles off its mortal coil, you can find me @jer_gib@types.pl on Mastodon.

@JAldrichPL You can't know for sure that it's AI. But you do know for sure that authors will tell you if it's not.

@jer_gib Not hard. Desk rejections are harder (how do you know for sure? hallucinated references are an obvious tell, but otherwise it can be hard)

So apparently AI use (mostly undisclosed) is becoming ubiquitous in ACM paper submissions. I don't think this happens in my group so I was surprised to hear it. But if that's actually the case, perhaps requiring disclosure is pointless. Thoughts?

@JAldrichPL My guess is that if desk rejections start happening, undeclared AI use will mostly stop. How hard is it to declare?

@jer_gib Under current ACM policies, yes. The question is whether at some point this becomes silly because everyone is using AI regardless. Not really a description of current reality in PL but apparently it is already largely the case in other subfields.

@JAldrichPL Shouldn't undisclosed use of AI lead to desk rejection? Chair's prerogative.

A notable issue is that ACM volunteers on the Ethics and Plagiarism committee of the Publications Board have to sort through policy violation reports. There's an argument that their time is better spent focused on cases where there is an integrity problem, not a disclosure issue.

@tritlo Search for "local handyman"? Or handyperson, of course.

wish there was something like a cleaning service but for miscellaneous minor house repairs (fix that light, this shelf etc). Just someone that comes over 2 times a month for 2-3 hours and does odd jobs like a rental groundskeeper

@headinthebox An AI company? Plagiarising? Who'd have thought!

amplifypartners.com/blog-posts/the… Pitched the idea of AI safety using proof carrying code to these guys on 5/10/2024 and they just ghosted me. Now they write a blog post how cool the idea of using formal verification for AI safetry is, with no attribution or acknowledgement. Not cool.

Do you have new research ideas in array-oriented programming? Then submit your work as a full paper or extended abstract to ARRAY 2026 by April 1 AoE! For more info about ARRAY 2026 and the submission process see pldi26.sigplan.org/home/ARRAY-2026.

@Gilad_Bracha Self-driving cars may indeed eventually put cab drivers out of a job. But the problem of 92-year-old Mercedes drivers accelerating into bakeries already has a solution.

@jer_gib Cabs cost a lot. AVs will eventually be much cheaper. Also cabs always as readily available as they are in large cities. AVs are the future; in some places, that future is already here.

Or, y'know, a cab?

German place-names rendered into English (morphologically reconstructed from historical forms)

@emax Sharika is trying to contact you...

If you want to say hi during my UK trip, I'll be visiting/speaking at the following places. Grab a slot on my schedule!

@BenSManning Think of it as describing your experimental methods.

Can someone explain to me the point of requiring AI disclosure on papers? It seems to be required everywhere I submit something... If I heavily use AI and there's plagiarism or fraud—that's on me. If I never use AI and there's plagiarism or fraud—that's also on me. Is the punishment supposed to be different? FWIW, I do disclose—I'm not trying to hide anything. I just don't get what it accomplishes besides maybe awkwardly, but not explicitly, discouraging AI use

Okay so, we just found that over 50 papers published at @Neurips 2025 have AI hallucinations I don't think people realize how bad the slop is right now It's not just that researchers from @GoogleDeepMind, @Meta, @MIT, @Cambridge_Uni are using AI - they allowed LLMs to generate hallucinations in their papers and didn't notice at all. It's insane that these made it through peer review👇

@haskellhutt @headinthebox They were somewhat overengineered...

@headinthebox It's perhaps worthy of note that the pyramids are still standing after 1000s of years, but houses built using modern methods are lucky to last 100 years :-)

By the end of 2026, we’ll look at Excel, Windows, Linux, Z3, ... the way we look at the pyramids or Machu Picchu: how on earth did the ancients build monumental software like that without AI?

@samth @heatherpedia I'll have to bite the bullet and replace our car soon - that will be the 4th in 33 years: 1991 Nissan Sentra (1992-1996, given to friends when we emigrated); 1993 VW Golf (1996-2010, given to a friend when a different friend gave us...); 2001 BMW 330i (2010-date).

@heatherpedia For me: 1989 Ford Taurus wagon (maybe my favorite out of all of these) 1998 Subaru Legacy wagon ~2005 Subaru Legacy wagon 2015 Subaru Forester (still have this) 2021 Chevy Bolt

I just bought a new car for the first time at 45. It's my 6th car in 29 years! 1984 Chevy S-10 1996 Dodge Neon 2003 Ford Focus car free for 2 years 2003 Toyota Corolla (by marriage) 2016 Honda FIT 2026 Toyota Corolla Cross

@tritlo @amezorr List comprehension do generalize: to ringads (collection monads). cs.ox.ac.uk/publications/p…

@amezorr to elaborate a bit, i’m always thinking about transformation of data. List comps work great for lists obviously, and even sets and dicts, but they don’t generalize. Function composition will always work

did a code interview thing Haskell was not an option, so I ended up writing a bunch of list(filter(lambda …, map(…,sorted(sorted(…,key=lambda …))))) the reviewers will have a stroke lol

@tritlo Now do data projectors.

there’s a negative correlation between the amount of computer scientists in a room and the likelihood of the audio working

okay, this is getting chaotic

We're delighted to announce that the JFP Special Issue on Program Calculation is now complete, and contains eleven papers that are freely available to read online! tinyurl.com/JFP-prog-calc

Torpenhow Hill in Cumbria supposedly means “Hill Hill Hill Hill.” Each part comes from a different language: Old English, Brythonic, Norse, and modern English. Basically, four groups saw the same hill and each decided to label it “hill” again.