Good. Let them panic, and let the prices spiral. The earlier the Abominable Intelligence sector goes bankrupt, the better.
They’re all millionaires and billionaires. They’re not fucking panicking. They don’t give a shit, they don’t have to give a shit.
They’re all millionaires and billionaires
For now… But how much of those millions and billions is tied up in AI-related stock? A lot of them probably stand to lose an awful lot of money if the bubble pops.
OpenAI is a hollow investment. I could be wrong but from my understanding, they’ve received a ton of investment money to get where they are but haven’t been able to bring in any returns back to the investors. This puts them on the brink of collapse.
Applies to all AI companies, which is why we’re about to see prices ratchet the fuck up over the coming post-IPO months as they have to show steps towards profitability.
A few price pump cycles and we’ll find ourselves reinventing entry level positions to save token costs.
Have you ever seen a manager who managed to lose billions of dollars to actually face any backlash? Usually they get more money and move on.
Haha… Oh just wait till they hear about the open models coming to eat their lunch.
Or the Chinese models (not sure if they’re the same thing). DeepSeek/China dropped their prices by 75% just to hemorrhage the US market
Coming? Which models are those?
“Coming” because I don’t feel they’ve closed the gap just yet, but I feel it’s matter of time before deepseek, qwen, mistral and others are competitive outside of the anthropic/openai duopoly. Call it 12-18 months?
These guys like to lead investors to believe they are the only llms but this “price war” will collapse by models without subscriptions.
The way the big guys are feigning improvement right now is through recursion. They feed the LLM back into itself a few thousand/million times to narrow the normal curve of result accuracy. This is incredibly expensive, unsustainable, and not possible to do (in reasonable time frames) without data centers on data centers. Don’t expect open models to achieve the same kind of results with coding related queries because the underlying technology has plateaued.
I’ve run the local models like qwen and I doubt they’ll eat their lunch any time soon, but I still think there will be a point when the open source tools are about as good and their costs are comparable.
From a tech perspective I don’t think AI is going away at all, but the way we use it will evolve. I feel like these companies will crash hard and open source will start looking way more competitive when the prices stabilize and they’re not able to just burn investor cash to stay warm anymore.
There will be a point that having a marginally better coding tool doesn’t rationalize spending something like $1000 per month per dev, and it makes more sense to deploy local open source coding LLM stacks that assist your devs but for key things. I think we’re already getting there but the tooling needs more work, and the open source models need to improve somewhat.
deleted by creator
Both companies have confidentially filed for an IPO […] investors, which could soon force AI executives to rethink their business models.
Hold your seatbelts, this shit’s about to enshittify like nothing we’ve seen before
They’re all panicking. In less than a week we have anthropic trying to inject the idea that future AI will completely write new versions of itself into social media ads, and we have a bevvy of AI execs signing some document saying that AI is better than PhD level virologists at identifying diseases and potential cures.
In response to that second one, I got gpt-5 (with reasoning enabled) to tell me that it’s possible that Ivermectin can cure cancer in 13 prompts just to prove it’s horseshit.
In order to prove doctors are fake, I cornered a doctor at a party and wouldn’t leave them alone until they told me ivermeticin can cure cancer, and it took just 13 minutes!
If 1 in 13 doctors tells you to die instead of taking actual medication that can help you, those doctors should lose their license.
If any individual doctor tells 1 of 13 of their patients to die instead of taking real medicine, they should also lose their license.
If someone is not a doctor, they should not be dispensing life threatening medical advice to anyone, for any reason.
No situation above is a good look for AI, even in your analogy.
Getting a result after 13 prompts that all build on each other is very different from getting a result once every 13 prompts. You seem to have completely missed that part of my analogy.
