• 0 Posts
  • 7 Comments
Joined 1 year ago
cake
Cake day: July 14th, 2025

help-circle


  • Yeah and the biggest issue is covered well here: https://www.notesfromthecircus.com/p/the-house-of-ellison-is-on-the-brink

    Essentially: the profits are in the infrastructure end (AI companies paying for parts to build the data centres) not the consumer facing end (the actual models - customer paying AI companies). The money is coming from investors buying in to AI in the hope of winning the supposed arms race to be The One True AI. The whole thing depends on the revenues in the consumer facing end going up to show this is a growing market and make people believe they will eventually reach profit. But the revenues in the consumer facing end are fake and being inflated by circular movements of money including companies like Nvidia paying companies like OpenAI to buy product, or OpenAI paying for product in stock. The revenue inflation is very similar to how Enron inflated it’s books its revenue appear higher than it really was.

    The brutal reality is: there will not be One True AI to Rule Them All. Instead AI is looking increasingly like a commodity - people will just use the cheapest, simplest tool to do the task, rather than play inflated prices for a swiss army knife AI. And people can host those cheap, simple AIs aleady at home, or on servers in the cloud, or companies can also do the same. The whole US bubble is predicated on spending as much money as possible to stake a claim in a future Google, when it’s looking more like it’ll be like the fast food sector with razor thin margins. Some of these companies may end up profitable but this is going to be a hell of a lot more competitive and less profitable than the valuations assume.

    Expect a very nasty correction.




  • Well the other side of that is managing software with hundreds (if not thousands) of person-years of development over it’s lifetime is also hard. There is a sunken cost fallacy with X11, and X11 was constantly extended to be what people needed it to be rather than the best design for the problems arising.

    Having said that, I’m not sure Wayland is the ideal solution. X11 doesn’t get in the way of solutions to problems, while Wayland really does. We needed a new and modern display server, but the Wayland project has decided to make that quite restricted - I personally think the security model is too far in the other direction from X11.