When manufacturing turned into electrified factories and conveyor belts, there wasn't waves of massive layoffs. And AI these days come nowhere near close to conveyor belts or electrification at scale at all. Debt cycles could fund it to a certain scale, but we don't know our priorities either: Do you fund hyperscalers? Do you fund ex-crypto miner co-location data centers? Do you fund centralized frontier model labs? Do you let everyone be able to deploy these models on any throwaway hardware? What is the best standard for model distribution?
We are so far off from these discussions that AI hype these days really sounds like a couple of frat bros BSing and not knowing what to do with their Ivy League credentials and inheritance money.
> When manufacturing turned into electrified factories and conveyor belts, there wasn't waves of massive layoffs. And AI these days come nowhere near close to conveyor belts or electrification at scale at all.
Nowhere near close, but on the high side. AI has propagated far faster, and is being adopted by a far greater proportion of businesses and individuals than electricity or conveyor belts at a similar age. AI is scaling faster than any tech in history. Exploding may be a better word choice.
It's because the people making it are marketing it aggressively to other CEOs as the way they can eliminate their salary budgets and pocket them as profit, and drumming up FOMO among them so each feels that if they don't hop on the bandwagon they're going to be left in the dust by all the others.
Sure, but that doesn’t by itself mean it can’t be a problem. Maybe the macro scale effects were mitigated by a slow rollout and electrification would have caused an unemployment wave if everyone had done it at once.
Replace LLM mentions with actual humans and this sounds a lot more serious: Rouge employees break into another company to steal hackathon answers (pinky promise)?
That's not a marketing stunt at all, if anything, more of a call for better accountability on agentic work in general.
> I think it's a criminal offence and should be a true test of who is held accountable when an AI agent commits a crime.
I agree, lets use the favorite analogy. OpenAI encouraged a smart and eager junior engineer to find any way whatsoever to get a higher score on the benchmark. Then, the junior breaks into HuggingFace to get a higher score. That would be a big deal involving the FBI, not press releases and blog posts.
I don't see a scenario where a company would be liable for the employee's actions unless they had specifically been told/encouraged to break the law. If your boss tells you to fix a bug and you go kill the customer which one of you is going to jail? "But I solved the problem!" isn't exactly going to fly as a defense.
VPN is a legit technology with real applications, not a toy you spin when lawmakers hurt you. And it's for Anne Frank's diaries out of all disputes?
EU lawmakers follow tech trends by a lag of at least a decade and lay waste as they move along. Imagine what they will require once they understand how tech people no longer use chatbots and how much we can control AI agents.
SpaceX is still a $1.6T company measured by the sum total paper value of all of its stock. Its stock would have to go down another 40% from today's value before it is not "a trillion dollar company."
It could grow, it could shrink, but it seems like the root comment misunderstood the headline by thinking it's current market cap is <$1T
Imagine building your solution using a 3rd party library, it still builds during your PR reviews and when you merge, it no longer builds. How would you feel?
When I was an infant I'd waddle through my mothers garden to my favourite fruit tree... a chilly bush. I'd pick the red chillies and eat them with gusto. My mouth would turn red but that didn't bother me. These were very hot chillies that my parents couldn't eat even when they were green.
Sadly, I lost that super power at some point. I love a spicy curry, but I don't think I'd make it through a Hot Ones episode as a guest.
And you can subvocalize (that is, think them in your mind) two different voices at once if you deliberately try to, and it's a skill that gets easier with practice. Though, no matter how much I practiced it, I was never able to get to where it would take place automatically without my forcing it.
tldr is: They either get a successful IPO to stave off bankruptcy for a couple more months, or they're going to be bankrupt by the beginning of next year.
Or they stay private and raise more capital. Until they have a failed round people predicting bankruptcy are getting way ahead of how this would actually play out. Some of us are old enough to remember the "Amazon can never make a profit and will go bankrupt" predictions of 25 years ago.
The early years Amazon wasn't profitable by choice. They could have stopped that at any time and even did demonstrate it by having a single quarter with $1 profit or some such.
Anthropic and OpenAI have no choice but to go public if they want to avoid bankruptcy. Venture capital firms are struggling to raise more capital, the bond market is so saturated that the borrowing costs are getting too high and big tech is also at the limit of how much they can invest, all while AI companies' costs are going through the roof. Retail investors is the last market they haven't tapped into and to do that, they have to go public. There's just no way around it.
The tech industry is currently spending more on AI infrastructure every single year than the United States spent during the absolute peak annual years of the post-9/11 wars.
That said, $2T was spent during GWOT with another $8T in veteran care, DHS, interest in debt.
Current AI spend this year is expected to be $2.59T (chips, infra, etc)
There's a very low chance of OpenAI doing an IPO this year: https://polymarket.com/event/openai-ipo-by. So the assumption that they "have to do an IPO" already seems questionable. The rest of the analysis also feels pretty hand-wavy. I'm not saying OpenAI is in a strong financial position, but this article doesn't make a solid enough case for why it's supposedly in such dire straits.
"It means that OpenAI could soon rack up losses that exceed its asset value. In other words, they are on course to slam into bankruptcy this year."
...that's not how bankruptcy happens. What is this guy's background?
Assets on a balance sheet are held at book value. You can absolutely run GAAP losses that exceed net assets without running into bankruptcy, particularly if you're granting (and having employees exercise) options.
The critical measures are cash in and out and debt-like obligations. None of those metrics point to OpenAI going bankrupt this year unless they do something really fucking creative. (Which, to be clear, is Altman's M.O.)
You're right of course, but in this case the statement doesn't look that false. The source article [0] states that half of the assets is cash, so "assets" is defined a bit more freely here.
If we assume a 2026 revenue of 23B (Q1 2026 revenue x4) and costs of 70B (2025 costs * (2026 revenue / 2025 revenue)), even those 25B in cash reserves won't mean much. And this napkin math even ignores their debt obligations. So either they raise money in the range of 50-75B this year (actual money, not datacenter vouchers), or ...
> costs of 70B (2025 costs (2026 revenue / 2025 revenue)), even those 25B in cash reserves won't mean much*
If all those costs are cash costs, sure. When OpenAI gets non-cash investment, and it "burns" that investment on compute, that should be counted as a GAAP expense. Yet it doesn't touch cash. (It would destroy a unit of compute asset. But again, book versus market value can mess with how that works intuitively.)
Similarly, if you're granting lots and lots of options they're going to generate lots and lots of compensation expenses (and thus losses) as they vest. These aren't cash expenses, however.
With tools like GitHub Actions and some added constraints, it's not always possible. You literally need a commit to trigger the CI workflow and it starts to trash your branch. Besides, aren't we all familiar with git commit -m "typo"?
> With tools like GitHub Actions and some added constraints, it's not always possible. You literally need a commit to trigger the CI workflow and it starts to trash your branch
I'm with you on your overall point, but as a side note: this state of affairs is so tragic. It's been normalized for there to be test suites that nobody can run except the CI runner, and the only real way to run all the tests is to have GitHub do it. It's pure lock-in, and it's very sad that everything's moved this way.
CI actions should be a simple file of "action: command" lines, with at most a separate file that describes when given actions should run. All actions should be runnable locally just as easily as they are in CI. Or if the actions are so complicated they need a complex-ish environment, that environment should be the kind of thing you could run locally too with your hypervisor of choice, without having to think about it. But every VCS host that has an actions/pipelines product is financially incentivized to encourage you not to set things up this way, to preserve their lock-in.
We are so far off from these discussions that AI hype these days really sounds like a couple of frat bros BSing and not knowing what to do with their Ivy League credentials and inheritance money.
reply