I had an interesting conversation at a friend’s birthday party with a few folks who were professionals but had been unemployed and looking for work for a while. I pointed out that with AI rapidly improving, many of the jobs that have been cut likely aren’t coming back.

They dismissed AI entirely as just a cheap imitation. Their experience was limited to trying ChatGPT over a year ago and seeing some clumsy early attempts by the New Zealand government to use AI. For them, that was enough evidence to label the whole field as an overhyped, short-lived scam.

It shocked me because, from my perspective, AI has been advancing incredibly quickly. I use these tools regularly in my work, and with a bit of focus on learning them properly, these emerging large language models (LLMs) are truly transformational. On top of that, innovation is accelerating rapidly, making AI both smarter and more accessible.

I’m not sure if we’ll reach AGI or ASI anytime soon, but it’s clear to me that society and our economy will be fundamentally transformed by AI.

This conversation reminded me just how much of a bubble technologists can live in. We see AI’s potential clearly and understand how quickly things can spread once they reach a tipping point. But most people probably won’t believe this transformation is real until it’s already underway. Instead of traditional economic institutions adapting their ways of working to integrate AI, we’ll likely see new institutions and methods emerge to replace the legacy systems entirely.

I’m genuinely concerned about how our economy will cope with the decoupling of work from primary economic systems. And when I think about how to spend my time while waiting for even more powerful AI tools—beyond just experimenting in my own work—I’m uncertain. Part of the answer seems to be designing new systems from the ground up around AI, and also continuing to tell people that AI isn’t just a passing trend.

This situation isn’t fundamentally different from what happened with Web 2.0 platforms like Twitter. The core human needs remained the same, but new technologies changed how we fulfilled those needs. Twitter didn’t replace our desire to stay connected with friends; it just made it faster and broadened our definition of who could be a “friend.”

So, looking forward, I think we need to ask ourselves: what would an AI-native version of everything we currently use look like? Most people and institutions won’t adapt—they’ll more likely be replaced. Does that mean we should just rush headlong into replacing everything with AI-driven alternatives?