Jacob Coxon's "whistleblowing" is a grift. The Hugging Face hack by OpenAI was because of horrendously and likely intentionally bad security practices and would have landed anyone else in prison. More real people are pushing back on data centers than even before. More companies are realizing that AI costs too much despite being massively subsidized, is horribly inefficient and even now, still largely produces shit results. To boot, open weight and open source models that cost far less to run are rapidly catching up.
So what's Big AI's solution to all this? Astroturfing in unison to say that they need to slow down to try to hide their collective failures and deflect blame when the bubble pops, which is looking more likely than ever and will be more economically catastrophic than ever.
I am tired of seeing this and the tech and finance media's continued utter failures to their readers to be critical and skeptical, instead providing an uncritical mouthpiece for the narcissist sociopaths at the wheel of this madness.
It's time for a good old rant as I rip this stuff apart and try to add a few more words of doubt into the continued circle jerk of the AI industry and the psychosis of their fanboys.
I'm not going to provide many more links in this post because any AI boosters reading will dismiss literally anything critical anyway. You can look up what I claim and find articles discussing it.
First, Jacob Coxon. The latest employee of Big AI who resigned because of supposed ethical concerns, yet is also cashing in on it with suspicious timing. It's clear to anyone who is interested in the truth that everything around his reveal stinks. He only joined Twitter (seriously, why do these people still use that shithole to communicate important things?) in January of this year and barely posted anything before his thread of poorly written PR-speak. But most telling of all, he not only had an exclusive article with the Wall Street Journal in place beforehand but that article was actually published before the Twitter thread.
Someone who quits a company for ethical reasons and wants to blow the whistle doesn't line up an press deal first. We don't know why Coxon left Anthropic and they strangely haven't said but I guarantee you it wasn't because he suddenly grew a conscience. This whole thing was staged, likely with a PR firm involved and not a single outlet who interviewed him pressed him on that or frankly, anything else because they all wanted the clicks and the interviews probably came with a condition that he get softballed. I'd wager that within a couple of months, we'll see an announcement of his new startup that will land millions in funding while producing nothing, just like the literally dozens of others from former Big AI employees and executives that have already done the same. If not that, he will either launch a Substack or some other paid media property. He didn't do all these interviews to just fade into the night. He either knew his time at Anthropic was limited or just saw an opportunity for a bigger payday and played the media like a fiddle. Hell, given what I'll talk about later, I wouldn't be surprised if it came out that Anthropic actually staged this themselves. Fuck Jacob Coxon and fuck the media that let him get away with this.
Next is the Hugging Face hack. Make no mistake, this is not a case of "rogue AI escaping containment". This was the result of security practices so weak and irresponsible that I can't see how it wasn't intentional. A competent junior help desk analyst will tell you that if you're testing a model's ability to hack something, you set it up in an airgapped sandbox with its target. There is no scenario where that model should be able to obtain access to the Internet and that this one could is a sign of either intent or layers of apparently some of the smartest engineers in the world being unfathomably stupid. In either case, this was a poorly configured experiment made by humans that has been spun to look like something else.
Hacking is a crime. I'm going to say that again because it's that important: Hacking is what OpenAI did and it is a crime. If any individual or smaller company did this, there would already be charges filed that would result in prison time and financial ruin. But when Big AI does it, they just go "Whoopsie!" and milk it for all the free publicity they can. This isn't just a failure of the media, this is a failure of the entire criminal justice system. Everyone from Scam Altman and Wario Amodei down to the managers and engineers who executed this hack should be hauled in front of a grand jury, forced to explain themselves and either charged with the crimes they committed or at the very least told they will be if this ever happens again.
I'll give you another example. Remember when Grok was being widely used to generate CSAM and non-consensual pornography? Not only did it take them two weeks to partially take that functionality down (paid Grok accounts can still make use of it to this day) but Elon Musk called government efforts to regulate it a suppression of free speech. Once again, the generation and distribution of non-consenual pornography or CSAM is a crime and all the images that were and continue to be distributed are still out there. What happened to xAI, Musk and his engineers? They received no consequences whatsoever and SpaceX was still able to do its rug pull of an IPO.
The AI industry has no business model and has no concept of a plan for one. The big players have made financial commitments so obscenely large that they literally cannot meet them unless they become the biggest companies in the world in only a couple of years. Titans like Amazon and Google have gone cash flow negative for the first time in decades and Microsoft is cutting every other part of its business to ribbons to keep their profits up while Satya Nadella keeps failing to make the AI play he staked his career on anything but an epic failure. Most of the AI GPUs you hear about in NVIDIA's record earnings are sitting unused and depreciating in warehouses because they can't build the data centers and scale the power grids fast enough. Yet they keep buying more and more each quarter, which is why NVIDIA is backstopping a frighteningly large number of data centers using accounting practices like Special Purpose Vehicles and circular financing, both of which should be illegal.
Every single query you've ever given an LLM has lost money due to massive subsidization and this will not change in the foreseeable future. This multiplies further because LLMs are still so inaccurate and error prone that things which should take one query usually take half a dozen or more to even make a functional result, much less a good one. Other than a basic search query, I've never had an LLM do something right the first time. Any other computing product that failed this often while also requiring you to pay for those failures would be laughed out of the room along with its creators.
Don't get me wrong, you'll still pay more as these companies slowly turn the screws and enshittify their products. Many business customers are quickly finding out. It still won't be nearly enough to make OpenAI and Anthropic sustainable though. Every new frontier model costs more to train and operate than the one before, even as the improvements continue to get more and more incremental. In any other real business, costs come down with scale but in AI it's the opposite. There is no path out of this that anyone has figured out, including the supposed smartest people in the world who are leading the industry. If customers had to start paying what it actually costs to make frontier models profitable, only the biggest companies and richest individuals could even consider using them regularly.
I actually think that because of this, we're eventually going to see a curbing of the tide of AI slop currently flooding the Internet because the subsidized costs make generating it either free or close to free. Once there's even a nominal price attached, it won't get used nearly as much for slop. That's just a guess but I think it's a good one and I really hope I'm right.
AI boosters will come out parroting all the talking points from the CEOs about the things it can do and glaze over its failures. Software engineers who want all the glory with none of the effort will caw about how great AI coding is, despite the code being inefficient, low quality, hard to read and often insecure. They don't comprehend that just because something compiles and runs, that doesn't make it good quality. A car with two cylinders misfiring and cement blocks in the trunk still drives but it's a shit experience. AI coding is even worse when measured in real unsubsidized terms because it often costs substantially more than it would to have a human write it better.
All of this has to come home to roost eventually because there is only so much capital in the world and a terrifyingly large amount of it has been devoted to this money burning endeavour that has questionable real value at best. No one in the industry, the financial apparatus funding it or the media machine fluffing it wants to admit they got it wrong. The reality is that Big Tech has run out of new ideas and is so used to being considered the sole source of society's continued evolution that they can't accept what they've become, a legacy industry that no longer capable of hypergrowth. It's run on hype, fairy tales, a near complete lack of accountability for its broken promises, investor FOMO and stock-based compensation for almost 30 years and doesn't know how to act like a real, mature industry.
We've now reached a new level of stupid where a grifter talks about how AI is going to kill us all while sidestepping any time he's he's asked to provide proof. Then almost to the day, Big AI CEOs are calling for an industry slowdown for "safety". It's not about safety, it's deflection. Their costs continue to rise faster than their revenues, their products haven't delivered a single one of the outlandish promises they've made and they're already hitting diminishing returns with each new model that four years into this boom, still makes constant mistakes and has no concept of efficiency. They're calling for governments to enact a slowdown so they can point at them when industry growth slows, the bubble starts to rupture and the grift catches up with them.
They're immeasurably arrogant but this goes beyond just ego. They all know the bubble will pop eventually because economic fundamentals say it has to. Thing is, OpenAI and Anthropic need to go public first so their founders, staff and backers can cash out their ill-gotten gains. The problem they face now is they've hyped themselves up to such ludicrous valuations that if they can't IPO at them, they'll get burned at the stake. It's looking increasingly likely that they can't and they're so desperate for something else to blame that they'll go so far as to say the technology they have made could kill us all if they aren't forced to slow down. Yet no one in the media is asking the obvious question: Why do you need to be forced to not make something you claim is so dangerous?
This strategy doesn't seem to be working. Trump isn't listening and in a feat of irony is calling AI skeptics stupid and low intelligence because this grift is literally the only thing covering up the absolute ruination his incompetent policies have brought the US economy. Without AI, he has to face the way things really are and he can't have that, especially when he's already likely to be kneecapped in the mid-terms.
I'm sick of watching the tech and finance press continue to fail their readers by propping up this scam.
I'm sick of seeing the industry and their fanboys continue to talk about things like AGI and Recursive Self-Improvement as if they're real things and not myths largely perpetuated by people who are always trying to sell something.
I'm sick of seeing computing and electronics become increasingly unaffordable because of the RAM cartel while enthusiasts continue to line up to give money to companies like NVIDIA that now spit on them.
I'm sick of seeing game developers I love dropping like flies because all the industry investment has been diverted to AI.
I'm sick of seeing slop games, YouTube channels, music and more made with technology trained on stolen works and the people who release it having the gall to call themselves creative because they think prompting a chatbot equates to doing real work.
I'm sick of seeing promising technology companies with actual good ideas not able to get funding because they refuse to be AI first when they don't need to be.
I'm sick of seeing people I respect continue to uncritically spew the industry's talking points and wave off how just bad LLMs are at most things after four years because they once did something cool for them that they were too lazy to learn on their own.
I'm sick of being scared to be the AI skeptic in work meetings because any attempt to say that maybe it shouldn't just be rolled out wholesale without serious consideration is met with eyerolls. And worse, being thought of as rude when the same people later complain about the costs and I dare to say I told them so.
I'm so fucking tired of all this shit.
This bubble will pop. I don't know when but I know it will. If it somehow doesn't, I will admit I was wrong and take a serious look at how my skepticism was rooted. But when it does, I know almost no one who drove this hype will do the same and no one who led this grift will see any real consequences for it. Make no mistake though, I won't let them forget.
Come at me if you disagree but if you don't bring receipts, expect no quarter.
Please let this be the peak of the stupidity. I can't even fathom anything beyond.
