Want to wade into the rainbow-ridden surf of the abyss? Have a sneer percolating in your system but not enough time/energy to make a whole post about it? Go forth and be mid.
Welcome to the Stubsack, your first port of call for learning fresh Awful you’ll near-instantly regret.
Any awful.systems sub may be subsneered in this subthread, techtakes or no.
If your sneer seems higher quality than you thought, feel free to cut’n’paste it into its own post — there’s no quota for posting and the bar really isn’t that high.
The post Xitter web has spawned so many “esoteric” right wing freaks, but there’s no appropriate sneer-space for them. I’m talking redscare-ish, reality challenged “culture critics” who write about everything but understand nothing. I’m talking about reply-guys who make the same 6 tweets about the same 3 subjects. They’re inescapable at this point, yet I don’t see them mocked (as much as they should be)
Like, there was one dude a while back who insisted that women couldn’t be surgeons because they didn’t believe in the moon or in stars? I think each and every one of these guys is uniquely fucked up and if I can’t escape them, I would love to sneer at them.
(Credit and/or blame to David Gerard for starting this.)
J-R Conlin has left Mozilla after 15 years, and has some choice words for the company.
This is by an llm-boosting firm, so be aware that it’ll have a lot of marketing in it. It doesn’t say nice things about vibe code (presumably because the authors want to sell you a solution) but the numbers are interesting even so.
https://www.faros.ai/blog/ai-acceleration-whiplash-takeaways
A few choice snippets, none of which will surprise anyone here:
- For every code change merged, the probability of a production incident has more than tripled.
The incidents-to-PR ratio is up 242.7% as teams move from low to high AI adoption.
- Bugs are accelerating, not stabilizing.
In our 2025 AI engineering report on the AI Productivity Paradox, bugs per developer were up 9% as AI adoption grew. In this dataset, that figure has risen to 54%
- The most experienced people in your organization are being buried.
Median time to first PR review is up 156.6%. Average time spent in code review is up 199.6%. Median time in review is up 441.5%. The engineers with the deepest knowledge of the system are spending their most valuable hours unraveling plausible-looking code that should never have reached them in the state it did.
This is fascinating to me and has shades of Iris Merideth’s The Problem is Culture. In other engineering fields, if you had a tool that cut costs but caused a threefold increase in failures you would be looking at a massive scandal, probably because if this was structural engineering rather than software engineering you’d be looking at a new Grenfell Tower or Hyatt Regency Walkway from every other project that used this shit. From what I’ve been following I don’t know that vibe coding has directly racked up quite so literal a body count yet, but if this pattern holds (and I see no reason to expect otherwise) then it’s only a matter of time before someone fucks up something important.
Also the fact that the framing here doesn’t seem to treat this as an existential risk to the project of AI coding is fascinating. If you’re not producing stable and secure applications in prod then what in the actual fuck are you writing all that code for?
I think part of the issue is that historical software quality was an artefact of its time… if you can’t easily patch your released products, you need to work harder to ensure they’re functional. If the only way for people to learn about how your product works in the documentation you ship with it, the docs need to be useful and comprehensive.
The combination of software needing no guarantee of merchantability or fitness for any particular purpose and the internet rendered those pressures obsolete. Ship shit, fix later. Mass-scale a/b testing over past decade or two shows that most people seemingly don’t care if their software runs like absolute garbage, and is covered in adverts, and harvests all their personal data and the leaks all of it that wasn’t sold.
An incident-to-pr ratio that’s up by 250% is unfortunate, but it is not yet so bad that the end-users actually care enough to do anything about it, even assuming they can do anything.
The “Behind the Bastards” on Nicole Daedone’s California sex cult does have a lot of similarities to Ziz’s stories of CFAR events, and the CFAR founders’ hints at what went wrong. They mention cuddle puddles and a series of early figures who dropped out or were pushed out. Its not at all the vision I got ten years ago from the blogs and podcasts.
Billionaires have a new start-up, Objection, that allows them to “sue” journalists by “summoning” them to a “tribunal” staffed by chatbots. They targeted journalist Gary Baum with their first “lawsuit”, which provoked Baum to write about them for the Hollywood Reporter. Like all vampires, upon being exposed to sunlight, founder Aron D’Souza
threw a hissy fithas shuttered everything “temporarily”.“I can’t tell you how many billionaires and CEOs have called me in absolute tears about their lives being destroyed by one article.”
I would pay to see a billionaire or CEO in tears over an article.
And if stopping them were that easy, why hasn’t it happened yet?
D’Souza has observed that his friends are “the best little boys in the world. They all went to the fanciest universities and won all the prizes.”
and wiped out several cities in the processMy profile of Sackler, it turns out, was the first case to be brought before Objection’s tribunal, although the company told me there are now dozens in its virtual docket. “You’re Exhibit A,” D’Souza said, observing that the verdict on my work was part of the company’s soft launch: “Building software is hard.”
did they try to turn their first target into unwilling and adversarial beta-tester?
After we spoke, I awaited my verdict before the Objection tribunal in the Sackler case. None arrived. Eventually, the landing page was taken offline. I asked D’Souza about it. He explained that Objection would “hold off publishing any adjudications” until “a new major strategic partnership” was announced.
so it seems
(As a general matter, D’Souza questions the common journalistic practice of quoting “experts” as part of coverage.)
it does fit a pattern
I don’t understand what the point of this business is, except to grift off the aggrieved rich failsons unable to handle the horribly difficult work of hiring a PR firm to smear the people they’re mad at. At first i thought that it could be to create a formal ‘social credit score’ for journalists and integrate it directly with different publications to quantify how mad the ruling class is with a given individual, in order to discredit them or bar them from work or chill their speech, as D’Souza implies here:
One of my final questions for D’Souza — who told me he’d been in a slew of talks with media owners about his venture (“I’m coming to New York next week to meet all the big guys”)
but that sort of thing happens already. Nobody who seriously challenges power is getting hired at The New York Times or The Washington Post. That’s just a top down directive from the owners. What is the point of this? it’s staggeringly stupid. Just shit talk these people in your private Signal GCs, guys. Andreessen and David Sacks and Karp will be happy to help you compose a peevish Xeet or a lawsuit. stop being weird losers.
Special mentions:
Then, of course, there are billionaires and their heirs. D’Souza believes that “many journalists are more powerful than billionaires,” explaining, “I can’t tell you how many billionaires and CEOs have called me in absolute tears about their lives being destroyed by one article.”
god, journalism would be so much cooler if it could directly remove money from the accounts of the Idiot Rich. Alas.
“It’s only the top 1 percent who matter. These are the people who are going to be the value creators” when, in his view, AI soon completely transforms just about every aspect of economic life.
Honestly, I think D’Souza explains the business best:
Once Objection issues an adjudication, satisfied clients can pay an extra fee to promote the finding “so it engages with the disinformation as it spreads through social media,” D’Souza says. “What I know from the Gawker litigation, having dealt with not just Hulk Hogan but dozens of other parties who felt like they were aggrieved by the media, is that they actually don’t want a financial remedy. What they want is a moral victory. Most of them just want a PDF that they can send to their investors and their family which says, ‘I did not go to Epstein Island.’”
Questions answered by t-shirt, etc.
In D’Souza’s interview with the Australian newspaper, he explained why: “It’s only the top 1 percent who matter. These are the people who are going to be the value creators” when, in his view, AI soon completely transforms just about every aspect of economic life.
D’Souza continued, “Ultimately, what’s the last job? It won’t be knowledge work. It won’t be physical work. It will be interfacing between the physical and the digital worlds, and right now that frontier is journalism.”
Taken together it becomes incredibly transparent that the actual goal here is to transform themselves into a kind of priest-king class, exercising absolute authority on behalf of the remote and unfathomable god that they built. Just please pay no attention to who built the AI, who runs the AI, or where all the money and power end up.
Twitter really broke their brains.
oh lol
Saw a guy watering his lawn this morning. Just standing there, hose in hand, dumping potable water onto grass that exists for no reason other than to be looked at and complained about.
Sir. Do you understand that a single hyperscale data center can drink millions of gallons a year keeping GPUs from cooking themselves while they generate a poem about a sad robot? That water has a HIGHER calling. That water could be evaporating off a cooling tower in service of someone’s RAG pipeline that returns the wrong answer with tremendous confidence.
And here you are. Hydrating Kentucky bluegrass. In a region where the grass was never supposed to grow in the first place.
I asked him if his lawn had an SLA. He said no. I asked what his lawn’s uptime commitment was. He looked at me like I was the unreasonable one. Meanwhile that turf is sitting at four nines of being green and producing exactly zero tokens per second.
We are pouring concrete across three states to host inference workloads, and this man is allocating municipal water to a crabgrass cluster with no monetization strategy. No usage-based billing. Not even a freemium tier.
Anyway I reported him to nobody, because there’s no one to report him to, which is honestly the most damning part of this entire ecosystem.
Touch grass, they said. He did. Look where it got us.
NOT EVEN A FREEMIUM TIER. that got me.
this man is allocating municipal water to a crabgrass cluster with no monetization strategy
This is poetry, AI could never
Coworker got fired because he used AI to plan for a site installment of our product. The AI made a very nice looking plan but it failed to include enough packing material so nearly half of the units arrived broken. Boss still thinks AI is going to revolutionize work for the better though.
to be fair, the plan would have worked if you had included the allotted number of goblins in the box in addition to the packing material.
Came across this today: the purpose of a system is what it does. Based on that I would say the purpose of Claude was to make mistakes and get my coworker fired.
UK government wants to mandate client-side scanning to stop nonces. You won’t be able to take or share nudes with hardware you own unless you submit to some dodgy third party identity check service first.
Unfortunately clippy the magic nonce detector does not exist. All this does is give our next Prime Minister (probably that froggy Farage bloke) totalitarian overreach
There seems to be a misunderstanding in that thread, not that the actual proposal is much better. Clippy isn’t expected to determine the age of the subject of an image, just whether the image contains nudity at all (in practice, usually how much bare white skin is in the image). Then, before your device allows you to take a nude photo of any kind, accept a text from your partner, or view a Renaissance painting online, it has to verify that you have a government-issued cybersex license to turn the filter off. For the children, of course.
Judging by the current state of NSFW filter neural networks, I expect a surge in the popularity of novelty color filters for smartphone cameras, racialized porn categories, and furry art. Online grooming focused on niche enough fetishes will likely be totally unaffected.
Finally, a machine which makes it impossible to watch the movie batman and robin.
(Joke explainer: the batsuit had nipples).
massive bong rip musk is broadsides-ing the spacex ipo so hard not only because he’s desperate for cash (he is) but also because he wants to stick it to saltman after losing the recent court case
-
"The instructions say “You are an agent that writes the various article sections for an encyclopedia entry on Joseph Kanuku.”
-
“[Note: Britannica is encyclopedia, but instructions say no Britannica. Wait, replace with another.]”
-
“Wait, instructions say NEVER cite Wikipedia. Oops. Let me adjust.”
-
“The instructions say prioritize peer reviewed, books, etc.”
-
“Wait, but instructions say never cite social media, so omit that last part. Wait, adjust. Since Instagram is social media, omit that.”
-
“Wait, instructions say avoid “References” as a section, but for completeness, I’ve omitted it from structure.)”
-
“The guidelines say “Include any of the following where relevant: - Factual details…” under Missing Information or Knowledge Gaps Examples. And in format, it’s to list only critical issues or all missing info or knowledge gaps.”
-
You know how the guardrails in the new anthropic models have been panned as overly sensitive?
Well now malware authors are including biology terms like “saccharide” in their source code to make llm powered scanners refuse to scan the code and let it past anyway
What a time to be alive
Amazing stuff. LLM security is so cooked.
The first I heard of a project doing this was iocaine, an anti-scraper tool (that I think awful.systems uses??) which has module names like “sex_dungeon” and “GargleBargle” intended to make the codebase unmanageable to navigate for llm tooling. It’s pretty funny that this has become an established technique.
awful does indeed use iocaine! and yeah the codebase is fun :)
it’s stunningly effective at handling scrapers, and it’s even more fucked how intense a plague the scrapers (combined with running shitty job scheduling with even worse code through resproxies) are
ig chucking in a segment of merck index (short entries with compound properties, 2000 pages of) to iocaine training corpus would trip it
Everybody remember Frontiers, the publisher that brought us the rat dck pck? Well guess what…
I’ve officially resigned as Associate Editor for Frontiers in Systems Neuroscience. It used to be a reputable journal, but became a case study in how forced automation destroys academic integrity. 👇
When Frontiers started automating the editorial process, I stayed. I reasoned that as long as the automation could be turned off, human editors can still ensure rigorous, high-quality peer review. This now became impossible - the system has been entirely hijacked by algorithms.
Over the last month I saw that human editors are now stripped of control. I could no longer stop the system from auto-inviting “reviewers” with zero relevant expertise. Even worse - the AI began actively revoking the invitations I manually sent out to actual, qualified experts.
I emailed and met with the editorial office to ask for the AI assistant to be turned off. I was told this is not possible. Instead, I was treated to some vague promises of potential future improvements and a dose of gaslighting.
If human editors can’t control who reviews science, it’s no longer peer review — it’s a rubber-stamp machine designed for volume and profit, not quality. I have no intention of attaching my name to it. So I’m out.
https://bsky.app/profile/michael-okun.bsky.social/post/3mnxkxte55s25
A predatory slop publisher has turned to algorithmic moderation by clankers? A shocking turn of events, to be sure. They’re probably just trying to wring the last few pennies out of their operation before they meet their inevitable oblivion.
Sigh.
ACM will no longer require disclosure of the use of #generativeAI in writing papers… an absolutely terrible change by the @ACM Publications Board to the Policy on Authorship. Yes, this will give ACM more papers. No, the ideology of “number go up” productivism is not the path to responsible, ecological or ethical computing, but to irrelevance.
I signed up for ACM last year, just hit my expiry (like, today), and truly don’t know if I want to renew
it’s been an extreme avenue of generative hype, constantly pushing talks, books, etc. I have no idea how they square it with the ethical pledge
By not thinking about it, one presumes.
I was flipping through this month’s magazine on the bus, don’t worry, they have a solution for reviewers…
https://cacm.acm.org/research/from-volunteerism-to-duty-reforming-peer-review-with-tokens/
:(
To prevent fraud and rule exchanges, tokens can be implemented as a digital entity, representing a fungible unit of value. Tokens can be transferred through cryptographic protocols and governed by rules running in the TMS.
They’re gonna do it. Peer review is going ON THE BLOCKCHAIN
Anything to avoid paying people for their time.
I wanted to give you a high five for telling APS where to go. That was rad as hell.
Link for that development, in case anyone missed it:
https://awful.systems/post/8263538/11477939
I did, indeed, turn them down, and I told the journal why.
Old: AI Winter
Bold: AI Ragnarok
Botterdamerung
Noice
AI Winter but with the same connotations as a nuclear one
The new Claude model will silently decide whether what you asked it to do is in line with anthropic ToS and silently corrupt your prompt if it doesn’t like what you’re asking. It’s couched as a “safety countermeasure” but it is presumably to stop Chinese labs trying to scrape synthetic data.
We’ve moved from ‘accidental’ hallucinations to deliberate misinformation and you’re paying $$$ for the privelige.
Claude can now be silently nerfed. Anthropic has decided it won’t tell users when this happens.
considering how many habitual llm users can’t tell good from bad output anyway, they always could have done that
To what extent should one trust a statement that a program is free of Trojan horses? Perhaps it is more important to trust the people who wrote the software.
Ken Thompson, Reflections on Trusting Trust
I know this outcome was inevitable after software became a mass market thing, but it’s still rather depressing.
“You’re holding it wrong” as official policy
Alamo Drafthouse built a reputation on strict viewing rules to provide a pleasant immersive experience at their theaters.
All of that is gone. They switched to you using your own phone to order food/drink so people are on the phones more often than a regular theater. And now they are doing AI “audience immersive presentations” where the audience remains on their phone to submit prompt garbage to AI generate dumb movies.
Support your local theater. This chain got too much love the last decade. Being in the northeast we only recently got an Alamo but plenty of small local theaters exist in and around the city (brattle, coolidge, west newton all if you are in Boston).
The slop startup dubs this “Audience Intelligence” and claims pixar experience. There are two “interactive movies” by them, Pickford AI. No employee there should even consider themselves adjacent to artists.
was there an ownership change or something? this is atrocious
Sold to private equity a few years ago and then purchased by Sony most recently
I have some modest proposals for handling private equity but they would all probably count as fedposting. We still haven’t found a decent replacement for the market niche JoAnn fabrics occupied.
New drama just dropped: The US government has forced Anthropic to block access to Fable and Mythic for non-US nationals, so they’ve blocked it for everyone.
They’re claiming they believe it’s because someone has managed to find a “minor” vulnerability, but if that were the case, the directive specified non-US restrictions, which makes little sense for an actual security exploit.
My assumption is this is part of the hasty xenophobic and protectionist policy the US is particularly fond of right now, but the government has failed to account for the multinational nature of the companies using them. I anticipate this restriction will be partly amended to accommodate that issue after they try to extract a pound of flesh from Anthropic.
I question how robust Anthropic’s geoblocking and data residency infrastructure even is. Their data residency info is littered with caveats that would make an EU market regulator shudder.
The move to block it for everyone does conveniently feed into Anthropic’s “it’s too powerful” narrative, but Anthropic is keen to demonstrate in this case that the issue they believe has been raised also applies to OpenAI, so I don’t think this was part of their original marketing strategy - even if it can easily be folded into it.
I think it’s more likely Altman is behind this as part of their IPO strategy. There was other murmuring this weekend about OpenAI considering drastic price cuts to compete. It’s an IPO race to the bottom.
It looks like AI use is quickly becoming what I assumed it would be, a weapon of the rich. The question for me is whether that weapon is a footgun, or has a much larger blast radius.
Yall sneer, but I made a couple of thousand bucks this saturday standing on a dark street corner selling illegal Fable and Mythic linkups. Gave the first question free to any boomers that came asking, to get them hooked.
Now seeing one-shotted Europeans calling for the development of AI/LLMs at home - “Mistral” is not adequate
Literally just started a discussion on European booster fanfic hype a few days ago: https://awful.systems/post/8591627/11736760
I wonder if Europe 2031 will get a boost out of this, it is really perfect timing for them, they can even claim an early prediction success on the US cutting Europe off! (But as with AI 2027, Europe 2031 assumes a much more competent US that can implement strategies like that in a competent fashion instead of some disorganized demands after 5pm on a Friday).
I’ve already seen “Europe 2031 was right. Shortest time between a work of science fiction and science fact” on LinkedIn… 🙃🙃🙃
The move to block it for everyone does conveniently feed into Anthropic’s “it’s too powerful” narrative, but Anthropic is keen to demonstrate in this case that the issue they believe has been raised also applies to OpenAI, so I don’t think this was part of their original marketing strategy - even if it can easily be folded into it.
It also saves them on the cost of actually serving the model, and stalls the cycle of people gradually realizing the new model isn’t much better than the previous one.
Very fair point. I’m still inclined to think this wasn’t primarily a marketing decision, because I don’t think they would point the finger so directly at the government if that were the case. Perhaps the directive was a blessing in disguise for them though.
Either way, I think this is pretty telling for what we can expect over the next couple of months before we see any public SEC docs from either OpenAI or Anthropic. If it weren’t for the immense damage it will cause for innocent people, I might have broken out some popcorn. Popcorn might be my staple food for financial reasons soon anyway.
