[256BitChris]

I could see OpenAI hitting financial issues which triggers some media induced panic and for people to claim the AI bubble has popped.

However, the core utility of the best AI (read: Anthropic's ATM, by miles), will still exist and be leveraged by those who have learned to use it well.

I could also see the exponentially declining power requirements offsetting the exponential-but-slower rate of AI compute demand, which then renders a lot of unused capacity in these massive data centers.

I think of it like the old mainframes in the 70s which would take an entire city block to run, and now we have the equivalent of millions, if not billions of them in our pockets.

[baq]

Anthropic isn’t the best by any reasonable measure. They’re the best in some areas and get pwned in others.

In general AI is very much like human intelligence in the regard that no two models are the same just like no two people are the same. IOW if you are a single model shop you might even not have any idea that you’re falling behind.

[jqpabc123]

I think of it like the old mainframes in the 70s

I think this is a good comparison to current AI.

billions of them in our pockets.

AI in your pocket (but first on the desktop) is a real possibility.

[cmrdporcupine]

The coming months are the reckoning in which the poor quality of the tooling and the safeguards around them become evident and hopefully eventually rectified.

By which I mean the competent organizations are the ones that will come up with cultural and technical solutions to manage the quantity and quality of the code better.

Others will suffer severe quality issues. Not because the "AI"s produce inherently inferior code but because the volume of the code is too high to manage review of, and to have good internal organizational knowledge of to manage the pages in the middle of the night when servers go down because of code nobody really understood.

I produce masses of independent project work all day long in my spare time using these tools and they blow me away. But in the context of professional work on teams of other coworkers the results are difficult to reason about and often impossible to competently review and it's not clear the results are superior. ' IMHO companies that drink too deep from the well without caution could be burned badly.

Aside:

I hate to say it, but there is no sense in which Anthropic has the clearly better product than OpenAI at this point. I know Claude caught developer's hearts through the fall, but GPT5.4 is a more powerful, careful, and competent model for coding and Codex is a far less buggy and more performant TUI. For the last 3 months I've gone back and forth between the two and I always run anything written by Claude Opus 4.6 by myself and my coworkers through Codex for review and it is constantly finding severe correctness issues to the point where I simply won't subscribe to Anthropic's product anymore.

On top of that, OpenAI provides far higher token limits. Even their $20 plan goes quite far.

If I was just building crud websites, probably Claude Code would be fine, and it does indeed show more "initiative" and "imagination" but I've seen it build way too many race conditions and correctness issues to trust it or the work my coworkers make with it.

[_puk]

A lot of anthropic's recent improvements are coming from the task focus and improved orchestration around the models, not purely massive changes in the models themselves.

This bodes well for us being at a point that even if the bubble burst, we'd still have usable AI going forward.

[eieje]

It’s pretty much undeniable at this point that the sentiment has changed.

About 2 months ago this place was unbearable - filled with doom and hype AI posts. I welcome the calming and eventual slow release of the bubble.