The less competent lean on the more competent, and soon the most competent thing in the room is AI

September 9, 2026

One way to think about the existential risk from AI is to realize that the less competent tend to end up relying on the more competent.

Think about the workplaces you’ve been in where there’s one or two competent people and like 10 lazy or incompetent people, and over the course of 2 to 5 to 10 years, you start to notice that it’s actually those two people who are doing most of the work and the managers just keep piling more and more onto them.

It’s hard for them to go on vacation, and when they complain, the managers don’t really help that much because they don’t have anyone else to give the work to.

Now imagine this for all of human society. The world runs on a whole lot of technology that very few humans actually deeply understand, know how to run. And over time, those people get older and retire. And so the list of technologies that are critical to the planet, where there are no humans capable of running that stuff, continues to increase.

So this transition could happen as some sort of abrupt or extreme takeover, but it’s even more likely to just happen slowly and naturally as humans simply lose the knowledge required to maintain anything that is running.

This is where the risk from AI starts to get very serious, especially as the AI gets smarter and smarter.

It’s not even necessarily a malicious thing, like we don’t even have to imagine evil in this equation at all.

It is simply very likely that in 5, 10, or 20 years, the world is mostly run by AI because the technologies that run it are too antiquated or too advanced for humans to know how it works.

So if we want anything, and if we need anything, or if anything breaks, we only have one recourse, which is to ask the AI to help us.

If you are reading this, I don’t have to tell you how dangerous this is.

All this to say that you don’t need to be a Terminator or Skynet type person to worry about the dangers of AI at an existential level.

We could have a completely benign AI situation in 10 years, where AI is running much of the world in decentralized or even centralized systems and everything just kind of works because somehow we got alignment right.

But when there is that much control in the hands of AI, it massively increases the risk of a much smarter AI and potentially a malicious AI just walking in and taking over the whole thing because at that point it will all be orchestrated in a way that it understands and that we do not.

These risks are not all science fiction. They are about to be very real and very practical.