Mastodon Feed: Posts

Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
ada@nyan.network ("ada f64.0") wrote:

Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
ada@nyan.network ("ada f64.0") wrote:

glyph ("Glyph") wrote:

RE: https://mastodon.social/@glyph/115647478978972835

ah shit people are faving this again, what fresh hell awaits me when I find out why

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

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

As a longtime collatz conjecture enjoyer, it is pleasing to hear that it proved powerful enough to resist proof by revealing a bug in the prover.

Boosted by jonny@neuromatch.social ("jonny (nonvenomous)"):
mevenlennonbertrand@lipn.info ("Meven Lennon-Bertrand") wrote:

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 provides a sorry-free proof of the Collatz conjecture, with llm involved
- the proof is detected as actually being a kernel bug
- 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) reproduce the bug, because they essentially copied the reference kernel implementation

And so
- adversarial AI writing proofs raises the bar for kernel correctness by a lot
- 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 catch logic bugs

isagalaev ("Ivan Sagalaev :flag_wbw:") wrote:

Summer day on the beach. This is how we roll here in Pacific Northwest :-)

#darktable #photography

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

giant hugweed

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

You know, I was in my 40s before I really heard of the concept of "resting" your food before cutting into it and honestly it still feels a little suspicious to me

Boosted by jsonstein@masto.deoan.org ("Jeff Sonstein"):
AaronDavid@beige.party wrote:

Boosted by jsonstein@masto.deoan.org ("Jeff Sonstein"):
AaronDavid@beige.party wrote:

jsonstein@masto.deoan.org ("Jeff Sonstein") wrote:

cool, a pair of red-winged blackbirds has shown up at the feeder I maintain in the backyard

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

Let me be clear: I don’t make things with generative AI, and I don’t endorse the technology. What I’m saying is that if someone else wants to use it, that’s not something that I’m going to hold against them. They have their reasons. Some of those reasons are actually good, even in the face of the broader problems. But it’s OK. We can still be friends.

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

I’m not going to die on the generative AI hill. I find the technology problematic, the companies that produce it reprehensible, and their leaders vile. I detest the way the models are made with stolen content, the wage suppression and job displacement, the environmental impact, and the cognitive decline. But I can still be friends with someone who uses generative AI, just like I can be friends with people who eat meat, and drive, and who have different politics than me.

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

This is not an endorsement of ambivalence; it’s me encouraging you to think about this and make an intentional choice. We’re all going to draw our lines in different places, and that’s OK. But you should always be honest with yourself, and be intentional about the (figurative) hills upon which you’re willing to die.

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

The person to your left voted for the other candidate.

The person to your right eats meat, and you’re a vegan.

The person behind you drives a car, and you take the bus.

The person in front of you uses generative AI, and you don’t.

You can spend your time focused on these differences, arguing about the ratio of harm to benefit. Or you can spend your time focused on the things we have in common, and enjoy being with others.

It’s a choice.

isagalaev ("Ivan Sagalaev :flag_wbw:") wrote:

Flight. The only passable photo of a bird from my outing.

#darktable #photography

glyph ("Glyph") wrote:

RE: https://infosec.exchange/@vathpela/117004523122718442

yesterday I was thinking about how infosec seems to be the one place where LLMs are doing work that seems to be making a measurable positive impact, just by observing the output in terms of counting vulnerability fixes. today I am reminded that even in this narrow area they still produce staggering time-wasting costs as well

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

Oh, I totally forgot to follow up on this earlier. It was fixed pretty quickly!

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

It had deleted the code in question before I could screenshot it, and i am not exactly practicing git discipline for a slop experiment, but here is a screenshot of the problem and the revised comment the LLM wrote about it in case you think i am making this up.

Boosted by isagalaev ("Ivan Sagalaev :flag_wbw:"):
karolherbst@chaos.social ("☃️karolherbst☃️") wrote:

To all #linux distributions out there: patch your libclc (especially the SPIR-V version) with this patch: github.com/llvm/llvm-project/pull/207264

Otherwise #rusticl is pretty broken (some applications e.g. #darktable won't work with it anymore without that patch, https://github.com/darktable-org/darktable/issues/21662)

Boosted by glyph ("Glyph"):
yossarian@infosec.exchange wrote:

we deleted support for three entire compression schemes in uv 0.12, and nobody has noticed (so far). I call that a win.

(Also a strong signal that Python packaging *is* - slowly - becoming a more uniform and standards-conforming ecosystem.)

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

Happy birthday @itswilwheaton. I have gotten you Crocs and a burrito.

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

I had never top to bottom vibe coded something before - like full on "do not look at the code, just keep rolling the dice" - and I was sort of confused about how people in forums would say they were "working for months" on vibe coding something. like what does that look like, what is the work like, what takes time, etc. So I have been trying it in the background when I take breaks from normal work, and while I have had lots of exposure to and used LLMs a lot before in existing projects, but this is the most profoundly frustrating and mind blowingly bad LLM experiment I have ever done and I am coming away with even less understanding of how people can think this rocks.

I had an old idea for a game I never followed through on, and the game is extremely simple (and actually pretty fun, unlike my other games): mechanical core is basically the game board is a graph, and moves consist of hops between the nodes.

The LLM was able to get a playable prototype running in like an hour, and I was like "whoa maybe this is good!" But none of the actual game part worked - there are invalid moves, but those were allowed. There were valid moves that were blocked, the game board was wrong, etc. OK, no problem, I can just refine from here. Just work through it and keep prompting until its done.

The moving target is unbelievable - every single change would bring with it new, unpredictable bugs. Since I hadn't been looking at the code, the LLM had made dozens of assumptions and hardcoded rules I never asked for and made no sense, and was re-reading those implicitly on every command so that internally it was fighting against me, trapped in a prison of its own wrongness.

Initially it handrolled its own graph layout algo which was terrible. I intervened and said "just use d3-force" since what I need for the game is at the level of something you could literally copy/paste from the docs examples. It did that, but the layout was still inexplicably bad - successive chains of nodes would hop back and forth across the whole board despite being connected to nothing else. Monday morning I started with prompting for a fix for the "chains of nodes always crisscross the board" problem. I checked in every few hours between work to find some new elaborate explanation of the problem, a few thousand more lines written, and claims it was fixed, and the problem being even worse somehow. Finally at the end of the day I got frustrated and broke the rule of the experiment and went and looked at the code - how could this be so hard, this should be a couple dozen lines.

In about 30 seconds of skimming, I found an elaborately commented function whose entire purpose was to guarantee that connected nodes were placed on opposite sides of the board. What??? There is absolutely nothing about what I wrote or could be implied from the structure of the game that would have suggested this as a thing to do. The nodes in the graph correspond to words, and the function specifically made sure that words that started with an even numbered ascii character were on one side and odds were on another. this was allegedly to "help with spacing between the nodes in layout."

It had been churning for a full day and generated thousands of lines of code to counteract a function that should not exist and makes no sense but was nonetheless clearly labeled as doing the exact thing I was telling it was a bug and should be fixed. if you told even the most inexperienced programmer to "fix the do x bug, the program should not do x," and in 30 seconds of skimming they found def always_do_x(), their first instinct would probably be to remove it. I looked at its tool use log and it had ingested this whole file, and this specific function dozens of times. The graph layout code was now like 3000 lines of spaghetti trying to counteract the function. I removed it and the problem immediately disappeared. Every single thing the LLM had said, elaborate, jargon heavy, math sounding excuses was 100% wrong, not even close.

The LLM dutifully made a memory about "always read the code to see if there is a function for doing the thing it is told not to do."

The gaslighting is mind blowing. Later, I told it to adjust the forces on the layout to make clusters clump up more and to spread out low degree nodes. This is an absolutely base level thing that force directed graph layouts do. This was fiddling with two or three constants. The LLM invented entire graph theory concepts and metrics to tell me that it was provably not possible to have less space around hubs and more space between them. It showed me series of fake fucking equations calculating some made up phrase like "unit density" in graph neighborhoods. IT TURNS OUT that it had just been redefining the force direction simulation in a LOOP such that for each node added, the ENTIRE GRAPH was excluded from the force direction calculation - only the newly connected node had any kind of layout forces applied and the rest were static. I removed that and it worked perfectly - as I knew it does, because I have done this a thousand times before.

I have never been more convinced that the driving force here is gambling addiction. I did indeed feel the addictive pull of just pulling the bar over and over and something happening. The code jittered, window refreshed... Dang not this time. Roll again. And when it finally sort of worked, here is the reward, this is amazing. I think that this game is good enough that ive been going in and trying to clean it up and make it actually playable, and credit where credit is due, it did program it so badly that I am now working on it out of anger in my free time rather than letting it languish.

Anyway, I wasn't planning on writing a blog post about this, but after its done and ive cleaned it up from the slop I will break out the raw message logs and write it up because this has been a completely perplexing look into the present and future of what people are telling me programming will work like forever now.

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

Pushthrough cat6a connectors that handle 23AWG cable ftw! Tonight I’ll run outside and cap off the other side of the buried cable, sneak in rev 3 of my elevation wedge, and pull the lose cat6 cable draped across the yard back inside so the yard guy can mow because it’s a jungle out there.

Boosted by andrewnez ("Andrew Nesbitt"):
githubchangelog ("Unofficial GitHub Changelog") wrote:

GitHub Actions holds potentially malicious workflows for approval https://github.blog/changelog/2026-07-28-github-actions-holds-unproven-workflows-for-approval
#github

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

Great meeting, and great after-meeting conversation too!

Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
zuthal@floofy.tech ("Zuthal") wrote:

you will now tilt your head in these directions individually

Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
inthehands@hachyderm.io ("Paul Cantrell") wrote:

RE: https://mastodon.me.uk/@Floppy/116997766777309382

This is a Hail Mary effort; hard forks like this are hard to sustain and rarely succeed. But it’s reached the point where we have to try.

Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
fzer0@nerdculture.de ("A Light Shining In Darkness") wrote:

If you see somebody stealing from AI data center construction sites you did not see somebody stealing from AI data center construction sites.

Boosted by cwebber@social.coop ("Christine Lemmer-Webber"):
wwu@social.wikiworkersunited.org ("Wiki Workers United") wrote:

Wikimedia volunteers have stood alongside workers throughout this campaign. Thank you!

Every movement is strengthened by the people who believe in it.

[union needed] #WikiWorkersUnited