AI Changes Everything–and Nothing

It’s not an exaggeration to say AI has taken the world by storm.

Remember this?

In early 2021 Mark Zuckerberg was on top of the world. Facebook stock, before its rebranding as Meta, had jettisoned to new highs from the depths of Covid unscathed. AI was not in the picture yet. The Metaverse and NFTs were marketed as the future. The assumption was people would buy digital assets, which would exist on the blockchain, or use VR goggles to interface with these virtual worlds.

Of course, those would go on to be among the biggest duds of the first half of the decade. Don’t feel too bad for Zuck–he’s still worth $100 billion–but the Metaverse is no more, taking with it $80+ billion of shareholder value. NFTs are worthless. Now it’s AI. Everything is computer. RAM has suddenly become scarce. Compute and tokens are the new oil.

I agree that AI has been transformative in many respects, but at the same time, much will remain unchanged. That’s why I position myself as something of an “AI incrementalist.” I don’t expect a crisis, mass unemployment, or a sudden “singularity” at some arbitrary date (which keeps being pushed outward when the awaited paradigm shift doesn’t happen). Instead, I anticipate steady sector-specific productivity gains, with most things staying broadly the same–just augmented by increasingly powerful AI tools.

It’s often assumed AI will have a societal leveling effect. I have long predicted the opposite: that AI will magnify innate differences of human capital. Smarter people will use AI more effectively and pull further ahead. Meanwhile, the gatekeepers of prestige (e.g. elite universities, top journals, etc.) will remain as exclusive and important as ever. Same for how AI will not lead to post-scarcity (healthcare, food, gas, tuition, daycare, insurance, cars, etc. more expensive than ever).

Same for how status will remain scarce too (e.g. top athletes, actors, and social media personalities). There is no AI replacement for the likes of Tim Cook, Tom Cruise, Messi, or Joe Rogan. 15 years later, online courses have not replaced Yale, MIT, or Harvard, and there is little reason to expect AI will either. AI cannot increase housing or make homes in existing expensive areas more affordable. Same for how AI has not appreciably hurt white-collar employment or depressed white-collar salaries, particularity FANG+ jobs and top quant firms such as Jane Street and Citadel, despite people having made this prediction since 2023.

For example, consider the prediction that AI will eliminate work. To the contrary, as the screenshot below shows, people still struggle with the learning curve (scroll to see both pictures):

The funny thing is, the “OP” was downvoted, presumably for having the audacity to ask for help when AI is supposed to make coding as easy as breathing. I also love the circular answer, “Input the error message into Claude to resolve the error.”

This agrees with I wrote in 2025 that AI only marginally reduces work, if at all. Instead, it changes the type of work. Rather than configuring LAMP, you have to configure AI runtime environments and repositories (e.g. GitHub, AWS, etc.). Coding libraries are replaced by AI libraries. Batch requests and command lines are just another form of coding. Projects are still hard and time consuming. Rather than eliminating coders, AI will repurpose them–shifting their focus toward debugging AI-generated code, refining and modifying outputs, setting up runtime environments, and integrating various AI tools (such as coding agents, OpenClaw, and model context protocols).

Billing has become very time consuming as well, such as managing token usage. Many are taken aback by surprise billing, compatibility issues, or arbitrary restrictions, which is especially problematic having already committed to a specific platform or model. These companies are not forthcoming about this.

True, AI has made some things cheaper, such as short-form video or image creation, but everything else where there is some form of imputed scarcity, status, or labor involved–is still expensive. An obvious example is advertising, which is booming, especially online, as there is no AI-based solution to create awareness about something. The fact that although it’s possible to generate near-infinite power or tokens, there are only a finite (slowly growing) number of sentient beings to consume the actual end product, means an irreconcilable bottleneck of attention.

A possible way to circumvent this bottleneck is the so-called “economy on a chip” model, where agents are somehow able to produce self-sustaining economies in their own right, trading with each other, as humans already do. But it would not be zero-sum, unlike daytrading or gambling, but would represent actual net-positive economic value creation that could technically count as GDP for the self-sustaining AI economy, which pays tribute to its human overseers.

AI paradoxically can have the opposite expected effect. For example, despite predictions of LLMs presumably obsoleting writers or writing, to the contrary, AI has led to a writing boom of the likes of which have never been seen before, due to four factors. First, AI assisting in the writing process, allowing more people to write competently. Second, productivity gains and wealth due to AI may mean more leisure time to read and write. And third, people writing about the very topic of AI, such as this article and thousands of others like it. For example, articles about “the implications of education and AI” or “AI and economic productivity” (or lack thereof) etc. And fourth, more platforms, such as Substack.

Moreover, old technologies have made a comeback, such as wired headphones, vinyl records, and typewriters. As it’s said, “Everything old is new again.” We’re going to the moon again. Pokémon, a staple of late 90s childhoods, is booming after a slump from 2000-2020. Collections being discarded or sold has led to a shortage of the most sought of cards. People are waking up to the fact that when AI can produce everything, some things, such as childhood memories, are truly irreplaceable. Or people try the new technologies and realize that they are not without serious limitations, like needing batteries for Bluetooth devices or reception issues.

Overall, nothing will fundamentally change, as is supported by the evidence. Three years since the AI boom began, things more or less are the same. People who predict some grand societal transformation towards more egalitarian outcomes are just projecting their desires.