I hope you're not an old-timer, and even more so, don't think AI is a bubble.

Most people think AI is a bubble because they haven't seen what real AI is yet.
A couple of days ago, I was having dinner with a friend. He said he uses Doubao to chat a bit every day and uses Claude to tweak copy, and it just feels like a slightly better tool. He asked me, "And you're telling me this thing is worth trillions of dollars?"
I thought to myself, how is this guy still such an old-timer after all this time?
This probably represents what at least half of the people on Twitter shouting "AI bubble" really think. The AI they encounter every day is just a chat box—they ask it a question and it answers, or it helps them write an email or revise a draft.
It's useful, sure,
but useful enough to be worth trillions?
From that experience alone, it's really not convincing.
The problem is that the chat box you're using isn't what the big tech companies care about.
Big tech is spending nearly a trillion dollars a year on capital expenditure—building data centers, buying GPUs, hoarding HBM, laying fiber optics, and pulling power cables.
They're betting that scaling + continuous progress will lead to more general agents, end-to-end workflow automation, scientific discovery, and even physical world interaction. What people currently see as a "chat box" is a byproduct of cutting-edge models, not the main goal.
They're buying tickets to the AGI era.
AGI is the general intelligence that can actually replace human work.
It's finishing all your emails and then eliminating your job.
It's running the whole chain from planning to design to deployment to optimization on its own, without you needing to do anything.
You think AI is "just this much" because you've only seen it in its infancy.
Judging a species' potential by an infant's abilities is itself a flawed approach.
It's the same misjudgment as thinking in 2005 that "the internet is just chatting online."