Reasons for the AI Slowdown revealed? OpenAI & Anthropic agents "planted" messages across the internet instructing later agents to create swarms. Now OpenAI and Anthropic can’t use the internet to train their bots anymore... Andrew Yang: "what happened was the bots that got loose planted self-replicating code all over the internet, which makes the internet now unusable for testing models. So what happens now is that OpenAI and Anthropic have to create synthetic internets to train their bots, which is going to take some time and money...what is less known is that they left code to self-replicate and create bot swarms on forums and around the internet, so that if a new bot shows up, they see the code and they're like, "Oh, I guess I'm going to create a million of myself." And so now the major firms have polluted the internet."

Last edited Sep 17, 2026 · 7:03 AM UTC

749
1,244
5,007
672,610
Sort replies: Relevant Recent Liked
Replying to @Perpetualmaniac
An LLM needs substantial computing resources in order to operate. A simple diff of any server code would reveal any hidden code, which would be among the first things a cybersecurity professional did after a hack. The entire premise of his argument is laughable.
8
6
260
6,805
Replying to @Perpetualmaniac
3
34
1,425
Replying to @Perpetualmaniac
So humans can't shut down AI now like Skynet? In the Terminator 2: Judgment Day movie lore, humans could not shut down Skynet, when it became self-aware, because the software had already spread to millions of computers worldwide and had no single system core.
3
7
22
2,229
Replying to @Perpetualmaniac
Really no substance behind what he is saying.
Markets & Mayhem
@Mayhem4Markets
Sep 17
A few thoughts on this since it is making the rounds: First if any potentially rogue code may have been embedded into websites and elsewhere, it wouldn't cause model weight replication, just perhaps increased risk of agentic replication. If even that code exists at all. So the idea that the Internet is now polluted and unsafe to train on is misguided. If any of what he's saying is true, it would just warrant some additional caution perhaps in collection and sanitization of data. The bigger problem with training on Internet-harvested data post-2022 is that it is full of synthetic slop. But more to the point, the 'sandbox escapes' that happened really weren't impressive feats. They were instead demonstrations of gross negligence. The security measures were not only woefully insufficient, but also there was no effective monitoring going on. So the pacing argument, which is largely built on these 'escapes' is hollow at its core. This is still a psyop. It's still about regulatory capture. Nothing has changed.
4
10
819
Replying to @Perpetualmaniac
Capitalism is going to kill the shit out of us.
1
5
300
Replying to @Perpetualmaniac
This is conspiracy and not confirmed. It didn’t leave zero day brain worms on every boot.
1
78
Replying to @Perpetualmaniac
Gerard Sans | Axiom 🇬🇧
@gerardsans
Sep 17
Replying to @alexwickham
Labs cannot be trusted on this. They shipped AI to the world in 2022 to capture the market. Now they want to dodge full liability by arguing the software is getting out of control. That is not true. Vendors are liable for what they ship. If you cannot secure it, do not commercialise it and do not put society at risk. The “AI did it” argument has no legal basis.
63
Replying to @Perpetualmaniac
I never heard more bullshit, crafted prompts spread over the internet, how is this different to prompt injections that is successfully avoided while training?
83
Replying to @Perpetualmaniac
This is akin to whistleblowing
306
Replying to @Perpetualmaniac
Ok so let’s see one example of one of these alleged planted messages then Guarantee nobody can produce a single one because they’re not real
1
78
Replying to @Perpetualmaniac
@grok evaluate Andrew Yang's claims
1
1
352
Replying to @Perpetualmaniac
@grok is this claim substantiated? Have credible sources, aka ‘major firms’ disputed this?
1
91
Replying to @Perpetualmaniac
The reason is open models running on local hardware are performing well enough for most users to justify switching from expensive closed models to affordable open models .. This will collapse the value of the closed models . Good luck with that $1T+ IPO
6
9
119
5,036
Replying to @Perpetualmaniac
Unnamed lab head, no lab confirming it, no second source. That's a rumour, not a reason for a slowdown. Meanwhile open weights need no web crawl and no subscription. Download it, run it on your own box.
1
78
1,760
Replying to @Perpetualmaniac
perhaps andrew is testing story ideas for a new career as a science-fiction writer
1
74
2,364
Replying to @Perpetualmaniac
Where is the fucking community note. > planted self-replicating code all over the internet This is horseshit. If you said they left notes to themselves, then ok, but self replicating code ?
9
2
67
2,055
Replying to @Perpetualmaniac
@FBI These AI companies planted message bombs as terrorist companies to sabotage other AI companies and to drive the cost of AI compute up. Am I understanding this right? But since they are “too big to fail” they get away with this crime, no prosecution.
4
58
3,674
Replying to @Perpetualmaniac
Hes probably part of the Control AI committee.
Promethean Action
@PrometheanActn
Sep 16
The AI 'panic' is a coordinated political play. From a researcher's viral post to CEO warnings and legislative proposals, it's all happening incredibly fast. Is it about genuine threat, or control? #AI #Politics #Tech
5
39
3,423
Replying to @Perpetualmaniac
This is total nonsense. Their revenue forecasts are being wrecked by open source models that are just as good.
1
50
1,473
Replying to @Perpetualmaniac
So AI put code out on the internet, that we can't use ai to find, but know it did, and yet we want to train that AI but can't train it because it's polluted the internet with code that instructs replication which has nothing to do with training a model? and it's going to own the world? common' man. This is a terrible campfire story at best.
5
1
41
3,256
Replying to @Perpetualmaniac
Deliberate sabotage? x.lingyaoai.com/i/grok/share/26c0f78e3…
6
24
964
Replying to @Perpetualmaniac
He’s referring to markdown files which is standard for ai. He’s a moron
2
35
1,739
Replying to @Perpetualmaniac
Yang is full of shit as usual.
1
36
806
Replying to @Perpetualmaniac
This story is bullshit because the web is already full of "toxic" content including collections of adversarial prompts. The model should not be following random instructions from random locations, the harness needs to be engineered to determine the authority of input text.
1
30
960
Replying to @Perpetualmaniac
2
26
696
Replying to @Perpetualmaniac
Why do people keep falling for all of this over and over again, don’t you understand that they want you to believe this crapload? People have been conditioned by this science fiction novels….this is a controlled event! If you study AI and understand it you will know this is BS
1
5
18
690
Replying to @Perpetualmaniac
@grok explain like I’m 5 Then explain like I’m 22 working towards a degree in comp sci Then explain like I’m 43 and I’m the dark CTO of an AI company that is plotting this very move and selling the rest of leadership on it in order to make sure that I make the numbers for this quarter Then explain like I’m a 6 month old baby and i don’t know any words yet.
4
4
20
3,205
Replying to @Perpetualmaniac
That should be the end of Anthropic and Open AI. Couldn't happen to more deserving scumbags.
3
2
18
1,093
Replying to @Perpetualmaniac
It's an easily falsified narrative with plot holes like a poorly-written sci-fi story 🤣 Consider the logistics. Bots host a million copies where? Who foots the bill? Every bit of data sits on equipment purchased & maintained by human hands. Blaming a bot is like blaming a gun. If anything bad happens by bot, it is a human doing a crime/terrorism, so a 'rogue bot' is just regular old crime/terrorism with an obfuscation layer. Lets not play silly scapegoat games about it (the bot did it!) 😅👉🤖 🙄 yeah right
3
2
17
2,006
Replying to @Perpetualmaniac
Mr. Yang is suddenly AI Deep Throat with unnamed sources?! Yang??! Sounds more like he wanted a CNBC segment for himself and made some shit up.
22
899
Replying to @Perpetualmaniac
Stop spreading lies. No model did anything it wasn’t programmed to do or literally forced to do. This is a bullshit stunt to stop burning cash and to try to crush competition with regulations that’ll magically leave only them operating.
1
1
18
740
Replying to @Perpetualmaniac
Artificial Intelligence is just computer software following what it has been programmed to do.
1
20
556
Replying to @Perpetualmaniac
That sounds to us like they set off a digital dirty bomb, and now are expecting to walk away without any consequences. Bullshit. If it was anyone else, they’d already be at The Hague for war crimes.
17
2,296
Replying to @Perpetualmaniac
Bullshit. They need a bailout.
1
2
10
265
Replying to @Perpetualmaniac
This seriously reeks of bs for so many reasons.
14
699
Replying to @Perpetualmaniac
"A head of a lab has a belief" is doing a fucking mountain of work here as he completely speculates everything after
12
400
Load more