Hello, it’s Rich, Aboard’s other co-founder. I want to call it now: We are in the midst of an AI bubble, and the crash is coming. Not that sort of bubble, or that sort of crash—though that may be coming, too. I’m talking about a very different kind of AI bubble popping. 

Let’s break down one of the most common and recurring narratives of humans collectively stepping in it—and then being shocked that there’s mud on everyone’s shoes:

1. It starts with genuine scarcity. Something that everyone seeks is just not widely available. Homes, credit, tulips, custom software, NFTs. The starting point is meaningful built-up demand. The result: Prices go up because demand outstrips supply. “Spending energy” is building up in the pipes!
2. Sellers notice the price signal and rush to provide “supply.” It’s often hard to manufacture the real thing, so there’s a rush to make the next-best thing. NFTs are a great example of this. While making your own cryptocurrency is challenging, anyone could create an NFT. The hunger was there. Give them timestamped pictures of apes!
3. Everyone gets confused because the newly available abundant thing sort of sucks. NFTs looked like scarce cultural assets, but they were inherently useless (and often ugly).
4. As demand starts to recede, the sellers know that the clock is ticking, so they flood the market with junk. While the window of opportunity closes, the sellers know full well that the party won’t last forever. So they flood the market with lower-quality assets while everyone is still drunk with demand. As a result, quality inevitably goes down. 
5. Finally, everyone bails in droves (i.e., a crash). When word gets out that the market got flooded with crap, the same momentum and velocity that got us here reverses—and the masses rip off their shit-stained shoes and run for the hills. 

Scale up the genius machine!

So much of the AI conversation today is about how to scale up all those genius coins. Anthropic’s Boris Cherny goes to ideas festivals and tells us how he’s been promoting some of his most talented bots into generals to manage his various battalions of agents. We’re constantly being reminded that without scale and velocity and massive investment, we can’t truly realize the value of AI.

And then there’s that intimate relationship between you and that prompt box. Have to write a cover letter? Want to file an HR complaint? Need an app for managing your collection of Funko Pops? AI will not only make it all happen, but it will always expand scope: It will reach further out beyond your own expectations. 

Having a conversation about that strange-looking scab on your elbow? It’s almost never a one-and-done. A quick answer is always turned into a conversation. You’re teased to keep going—to keep conversing. 

Dear readers, we are in the midst of step 2 in the outline I shared above. We are still in awe of the magic of these tools and the “sellers” are selling as much as they possibly can. But steps 3 and 4 are coming.

Essays meander and reek of unoriginality. Funny images are boring now. And a lot of software built with AI genuinely sucks. It’s often bloated, incoherent, and weirdly unnecessary. This is happening because our own judgment today is still clouded. When AI is used diligently and with careful thought, rather than complete delegation, it is transformative. But we’re not there yet.

We talk to a lot of organizations about their AI strategy. There are more than a thousand “head of AI” jobs in the markets. But everyone we talk to seems to be coming to the realization that quantity not only doesn’t equal quality, it is a liability. Used poorly, AI can generate an immense amount of technical debt at a remarkable rate. “Tokenmaxxing” seems to be dead. I’d argue that we’re in the “huge quantity of low-quality goods” part of the cycle—the “Paris Hilton sharing her NFT on Jimmy Fallon” phase of the bubble cycle.

Returning to scarcity

If the market is being flooded with badly written essays and lousy software—the equivalent of someone saying “look at this picture of a cool robot dog I paid $15,000 for”—how do we prepare for a post-AI bubble world? Because I do believe that, unlike NFTs, there’s real value here.

First, we need to turn away from all that convenience. We have to relegate AI to a tool that can enhance and grow our own creative skills and abilities, not supplant them. If everyone keeps delegating, the ability to differentiate and stand out disappears. 

Second, I would push people to return to the one asset that hasn’t been commoditized: Your craft. I’ve witnessed engineers and designers with 10-15 years of experience put it all aside to let Claude run for hours in blind faith. The result is not better; it’s almost always worse. AI must be subservient to our creativity, knowledge, and experience.

And finally, ask for less. Demand less from these tools. Give them way less surface area to work. Your prompts should be very long, and the leash should be very short. Trust yourself to have better judgment than the machine. 

The fever always breaks

Recently, Elon Musk was interviewed by The Economist and he predicted a future of abundance where everyone has more free time because AI has taken on most of the world’s workload. I don’t think that’s going to happen. Markets crash because humans are finicky, and while we do get drunk with greed when the bubble inflates, we consistently call out crap that seems to have very little value.

Musk’s prediction is wrapped in a kind of cynical optimism: He’ll keep making rockets and cars while the rest of us lie around on the rocks like fat seals, liberated from the burden of having anything useful to do. I don’t buy it. Humans have a remarkable capacity to get drawn in by abundance, but an equally remarkable capacity to eventually call bullshit. That’s what a crash is: the moment we stop confusing supply with value. And that’s why I’m optimistic about the one that’s coming.