"We can't ask AI, it lies" vs. "Here is my superpower prompt"

It’s well-documented that Gen Z / Gen Alpha are not exactly fans of AI1. And that older professionals can’t live without it.

My own 9 year-old kid has told me that games made with AI are terrible and that “we can’t ask AI, it lies” (in the case we were talking about, it did lie). Marc Andreessen believes he has a superpower prompt that helps him understand anything.

The easy conclusion is that the youngers don’t like AI because it’s taking away all the entry-level jobs. Like many easy answers, this only scratches the surface.

There are also some fundamentals baked into AI that give a fuller picture of why there is such a disconnect:

What you have in your mind when you’re dealing with AI

What they are trying to do, and how good AI is at the task.

What you have in your mind when you’re dealing with AI

If you grew up with Star Wars, Wargames, Terminator, Robocop through to The Matrix you’ve got a clear idea of what AI is. It’s something that has its own motivations. It can be helpful to humans (C-3PO in Star Wars) or something all-powerful and impartial (Wargames) or something that is very bad news for people (Robocop, Skynet and the Matrix’s Agents).

You’ve had decades of indoctrination about what to expect when you meet AI.

And here you are chatting online with something that you’re told is AI and which certainly feels very similar to having any other online chat with a friend or co-worker. Whether you’re Blake Lemoine in 2022 (the Google Engineer who thought Google’s LLM was conscious) or Richard Dawkins in 2026 who thought his chatbot was conscious or even Marc Andreessen who thinks that writing Never hallucinate or make anything up in a prompt is equivalent to giving an instruction to a human who works for you2.

Or whether you’re one of the hordes of people who say “I thought X and I asked AI and it said yes” and take it to mean that there is something genuinely agreeing with you that X is true3.

These are all smart people. Yet they see sentience where there is none.

The younger you are, the less time you’ve had to build in all these assumptions about what to expect when you meet AI.

What you’re trying to do, and how good AI is at the task

Large Language Models are the great averagizer. They are the average of all the content in the world4.

In my professional life, this is super useful:

In my own area of expertise, average is good enough a lot of the time

There are niches in my own area where the average of the rest of the world is better than what I can do

And outside of my area, the average of the world is a great starting point.

Examples:

Most web applications have similar patters for how users interact with them, so an average piece of code for making a database update is going to be fine.

I’d never written an MCP server but loads of people had. Claude was great at building me my first one.

I’m rubbish at job specs. Lots of people out there are much better than me. ChatGPT is a great start when I want to hire a new role.

But that’s because I’m pretty old and I’m settled into my niche. I have a pretty good idea of where I need help and where what it’s telling me is going down the wrong track. I know when it’s starting interfering with code that it shouldn’t be interfering with. When it comes to job specs, I’ve seen a fair amount of them in my life, so I can take what it started with and flesh it out more.

In contrast, youth is about possibilities, exploration and figuring out for yourself what your place is in the world. Youth is bound to react negatively to something that tries to steer it down one specific (average) path.

It’s also because “how to write a job spec” is a pretty well solved problem with millions of examples out there. If you’re trying to do something a bit more niche then the LLM is less likely to have a good set of data to access.

For example, my kid was trying to figure out something on a new Roblox game. I suggested we ask ChatGPT. He said “we can’t ask ChatGPT, it lies”. Sure enough, it hallucinated something based on a similar Roblox game. There are many Roblox games each with different quirks and you can’t generalize across all of them.

The kids are alright

At the end of the day Gen X (and probably older) people are in a double-whammy which makes it hard for us to judge AI accurately:

The feeling you get when interacting with it does seem like something aligned with the popular fiction we grew up with

Most of the time what you’re doing with it aligns with what it’s best at.

The kids, though, are alright.

Notes

As is usual for 2026, in this post when I saw “AI” I generally mean a Large Language Model used to generate text. It pains me that “AI” and “LLM” are seen as interchangeable terms, but we are where we are, right? ↩︎

I bet you that the responses he gets from AIs using this prompt still come with the “AI can make mistakes” disclaimer, even though he told it not to make mistakes. ↩︎

The tendency to act as if a chatbot is human is not new. See https://en.wikipedia.org/wiki/ELIZA_effect. It’s also the concept that underpins the Turing Test. I feel that this has only got more pronounced given all the branding around “ChatGPT is AI, like the science-fiction you’re familiar with”. ↩︎

Or, if you prefer, they consume huge amounts of text and use it to create one version of reality. In my mind, this “one version” is equivalent to “average”. ↩︎