Mastodon Feed: Posts

Boosted by glyph ("Glyph"):
davidgerard@circumstances.run ("David Gerard") wrote:

having real trouble finding actually useful work by alphafold

wikipedia page has a fuckload on it winning competitions and also it uh predicted stuff about the COVID virus that was already known

if there was anything it'd be there

anyone got links to anything convincing?

Boosted by glyph ("Glyph"):
jaafar@hachyderm.io ("Jeff Trull") wrote:

Following both @selfcare and @simple_sabotage is great because they are released simultaneously and produce insane juxtapositions like this one

Boosted by cstanhope@social.coop ("Your weary 'net denizen"):
fnf@mathstodon.xyz ("Fredrik Nordvall Forsberg") wrote:

The TYPES 2025 post-proceedings have been published!

https://www.dagstuhl.de/dagpub/978-3-95977-441-3

Due to the efficiency of the Dagstuhl staff, some extra work was created for them by having to change the publication date from August to July.

Thanks to all authors and reviewers for all the hard working going into making this an excellent volume.

Boosted by cstanhope@social.coop ("Your weary 'net denizen"):
mevenlennonbertrand@lipn.info ("Meven Lennon-Bertrand") wrote:

EDIT: I jumped too fast on the easy story, and overplayed the role of AI here. Had I known how it would blow up, I would have been more careful... Things are still not entirely clear, but I've tried to make the post less wrong.
---------------------

This whole Lean kernel bug is almost too on point to be true, it fits perfectly in the discussions we've had here and elsewhere over the last months/years…

To summarize:
- formal methods researcher (@ramana) sets up a repo with a sorry-free proof of the Collatz conjecture
- the proof is reporting (by @kirancodes) as being a kernel bug, the repo was a tongue-in-cheek way of exposing the bug
- Ramana mentions llms were
involved (but how is unclear atm)
- the bug is related to (nested) inductive types, for which there is no clear theoretical specification: the kernel's code is the reference
- external checkers (lean4lean and nanoda from a week ago) are affected too, either because they copy the reference kernel implementation, or because (nanoda) of a missing check in this subtle part of the code
- the bugs compound, and the invalid proof is accepted by comparator, Lean's gold standard for proof checking

And so
- adversarial AI writing proofs raises the bar for kernel correctness
- without a clear type-theoretic understanding of *what is actually implemented*, we're toast
- external checkers help to catch implementation bugs, but without a clear specification they can't really catch logic bugs

I believe the whole story also showcase the solidity of human measures: the bug was fixed as soon as found, with healthy discussions between experts, in the open, as to what to do. At no point was there any real doubt as to the solidity of a mechanised result.

EmilyEnough@hachyderm.io ("Emily 🏳️‍🌈🏳️‍⚧️") wrote:

I got out before the heat this morning and put a new connector on the buried gen2 starlink cable so it would work with the new dish, and was then able to clean up the temporary cat6 cable I had laid across the yard.

What took me 2 minutes inside sitting at my workbench took 45 minutes standing on top of a ladder in the yard, because I just could not get the cables straight enough to go through the connector in the right order.

But it is finished, and I should be done climbing up and down the ladder in the yard.

jscalzi@threads.net ("John Scalzi") wrote:

THE LITERATURE "AI" COULD NEVER - AND NEVER DARE - TO WRITE

May the universe bless Chuck Tingle and keep him always

Boosted by ChrisWere@toot.wales ("Chris Were ⁂🐧🌱☕"):
FediThing@chinwag.org ("FediThing :progress_pride:") wrote:

There are some really interesting videos on PeerTube nowadays. If you want a guided tour you might want to follow my account @FediVideo which boosts a human-curated selection of them.

#PeerTube #Fediverse

soatok@furry.engineer ("Soatok Dreamseeker") wrote:

I get some weird emails.

How do I break it to them that "Barkane Arts" is a joke organization name that I came up with as a play on the term "computer wizardry" (but with more furry) rather than an LLC registered in any state?

neatnik@social.lol ("Neatnik") wrote:

brb, removing the plank from my eye

neatnik@social.lol ("Neatnik") wrote:

Why shadowban when you can just hold people accountable with a clear, regular ban?

Boosted by neatnik@social.lol ("Neatnik"):
AltTextHealthCheck ("Alt Text Health Check") wrote:

Results from tracking 16,650 images across 1,016 instances from the mstdn.social federated timeline.

Then filtering down to the top 1000 instances by active user from instances.social who had posted at least 25 images (1.5x the average).

Boosted by neatnik@social.lol ("Neatnik"):
RhinosWorryMe@climatejustice.social ("Rhinos Worry Me") wrote:

@mtconleyuk @dbattistella

#Climate #ClimateCrisis #ClimateCollapse

cwebber@social.coop ("Christine Lemmer-Webber") wrote:

RE: https://mstdn.ca/@dbattistella/117004024280410659

It's worse, because this isn't a warehouse being filled and pulped once, but over and over again.

Boosted by brib@bribstodon.xyz ("brib :neofox_floof:​ :Nonbinary:"):
datarama@hachyderm.io wrote:

There's a deeply depressing picture going around of Anthropic's enormous collection of books that they shredded to build their anti-library.

So here is a picture of my collection of well-cared-for and -about books. I will add a picture of each subcollection in the comments - and please, *please* add a picture of your own book collection, if you have one. It doesn't matter if it's humble, as long as it is books that are meaningful to you and that AI companies *aren't* adding to the book burning pyre. I would especially appreciate if your picture comes with a little story.

andrewnez ("Andrew Nesbitt") wrote:

Working on a license scanner for git-pkgs, uses the scancode license rule corpus but much, much faster and less resource intensive: https://github.com/git-pkgs/licenses

My plan is to eventually use it to scan every version of every package with it soon.

dysfun@treehouse.systems ("gaytabase") wrote:

RE: https://lipn.info/@mevenlennonbertrand/116997917683191056

okay the lean llm proof thing turns out to be even more hilarious.

it was a jokey way of demonstrating the lean kernel bug. everything else was people who didn't read it getting carried away 😂

top inadvertent trolling skills tbqh. it's not even clear how much LLMs were used.

Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
dbattistella@mstdn.ca ("DB 🌱💦") wrote:

This is the magnificent library Anthropic amassed, before destroying it because it gave them a legal copyright loophole.

#Anthropic #AI #Books #Culture #Tech #LLMs #Civilization

Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
dbattistella@mstdn.ca ("DB 🌱💦") wrote:

Anthropic spent billions of dollars to slice the spines off millions of books, including very rare ones.

AI companies are purchasing large quantities of used and rare books and scanning their contents to train LLMs. They then turn the originals to pulp.

The sum of human thought, digitized and shredded. Every book that gets scanned disappears from the physical world permanently.

Books that survived wars, fires, and centuries of handling are being shredded so an AI can learn to write a better marketing email.

#AI #Books #Culture #Civilization #Tech #LLMs

Attachments:

• video: CU of an industrial-size blade slicing through the spine of one old book after another.

Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
PhoenixSerenity@beige.party ("Ms. Que Banh") wrote:

At Fisherman's Wharf, earlier 🙂 With @mayintoronto.
Photo by @kaosk

Boosted by cstanhope@social.coop ("Your weary 'net denizen"):
sharktastica@toot.wales ("Admiral Shark") wrote:

#SHORK 486 #Linux can now include cURL, Lua & Tilde! GNU Emacs & micro are also under testing and should be available soon, pending optimisation. A #writerDeck centric edition of SHORK can now be built, which prioritises including text editors and fewer distractions. https://github.com/SharktasticA/SHORK-486

Boosted by brib@bribstodon.xyz ("brib :neofox_floof:​ :Nonbinary:"):
Eichi@mastodon.bierschutzpartei.de ("Eichi🐺") wrote:

Can anyone recommend some basic (interactive) bash tutorials suited for people with next to no knowledge?
I'll have to teach a Newbie in about a month :p

Kennt hier jemand gute (interaktive) Tutorials für Bash Grundlagen, die man jemanden geben kann der kaum bis gar keine Erfahrung hat?
#Linux #linuxhelp

Boosted by brib@bribstodon.xyz ("brib :neofox_floof:​ :Nonbinary:"):
yaxu@post.lurk.org ("Alex McLean") wrote:

Thinking about earlier days of free software, I guess GPL was designed for a world where you might get a program on tape, floppy disk, or downloaded from a local BBS (probably historically inaccurate, but stay with me..). You might have an address to write to a human to get the source code off them, or find them and ask for it in-person. I guess the code would move slowly, getting improved in different ways depending on the community it reached.

In the corporate internet days, almost everything goes through a huge massively centralised, closed source (!) Microsoft AI platform, constantly scraped and IP washed by extractive bots, that's a totally different prospect and completely changes the meaning of the license, and sometimes feels like it makes the whole process of sharing feel inhumane.

I'm now thinking about no longer uploading copyleft source code to codeberg etc as well as github, only sharing it at workshops/with friends, and asking others to do the same. Maybe this could be part of the license, but could just be an unwritten social contract.

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
t_var_s@phpc.social ("Ricardo Tavares") wrote:

Programmers grow in knowledge as they write code. Compilers are deterministic and don't cost you more for each line of code.

Determinism and practical knowledge is how most systems are maintained. It's how you fix bugs.

LLMs don't learn anything after being deployed. They're probabilistic and cost you more the larger and more complex your system becomes.

Are we thinking of tests like we've never done before? Are we prepared to build platforms again from scratch when they become untenable?

Boosted by cstanhope@social.coop ("Your weary 'net denizen"):
kimcrawley@zeroes.ca ("Kim Crawley 😷 (she/her)") wrote:

Share! 🎉🎉

I'm just a former cybersecurity professor who quit her job because the Open Institute of Technology was going to force me to subject my students to Gen AI.

Kickstarter is my sole source of income for the time being, and I'm really struggling financially. I wrote books for O'Reilly and Wiley, and now I write self published books criticizing Big Tech.

I have an excellent track record with Kickstarter campaigns.

Gen AI will end. The laws of physics can't sustain datacentres. Financial bubble. Model collapse. I cite MIT, UN, and Harvard Science Review.

What will we do in the aftermath to recover from massive population psychological damage, environmental destruction, tech debt, abandoned datacentres?

Pledge as low as about $12 USD for an eBook.

Share! 🎉🎉 #NoAI

Kickstarter link:
https://www.kickstarter.com/projects/kimcrawley/recovering-after-gen-ai

andrewnez ("Andrew Nesbitt") wrote:

Wheels, Bottles and Images

https://nesbitt.io/2026/07/30/wheels-bottles-images.html

baldur@toot.cafe ("Baldur Bjarnason") wrote:

Starting to see posts and videos from dev/UX people along the lines of "AI works really well but for completely unrelated reasons I now hate my job, am feeling lost and burned out, and am planning to leave the industry"

Too bad nobody tried to warn them and got repeatedly ignored or dismissed /s
🤷‍♂️

fribbledom ("muesli") wrote:

Every technological breakthrough of the last 30 years eventually reached its final form:

More targeted ads.

Boosted by brib@bribstodon.xyz ("brib :neofox_floof:​ :Nonbinary:"):
vicorva@mastodon.art ("V, The Dragon Witch 🌙") wrote:

Hey, if you use Stripe for anything, just FYI that you have to *opt-out* of their new AI, which they will charge you for.

You don't want Radar Standard (which is free until January, I think, and is their AI plan), you want Radar Lite.

You can opt-out in your dashboard.

dysfun@treehouse.systems ("gaytabase") wrote:

one thing that i somehow never really noticed despite years of rust is that you aren't passing the ability to drop something or give it away when you pass an owned, you're passing the requirement to

pzmyers@freethought.online ("pzmyers 🕷") wrote:

When that evil crook dies, I'm going to want to raise a glass. Don't you dare stop me.

https://freethoughtblogs.com/pharyngula/2026/07/30/political-discrimination-against-beer/