My prediction on AI in the next 5 years

This article is my take on where the AI space goes over the next 2-5 years.

Not the model race everyone is watching.

But the layer being built on top of it and who will make money once it exists.

The short version…

Models will disappear. Not literally, but from your awareness.

You won’t touch it, you won’t think about it, and you won’t pay for it by token.

I’ll cover 5 topics inside:

1. Why models are commoditized
2. What will replace them
3. The model I want to see exist
4. The weird agent economy coming
5. Where humans will fit in

The model layer is getting boring.

More companies will provide small, medium, and large models.

One for every use case until they’re commoditized.

I already feel like LLMs are plateauing in real time…

When I compared Fable 5 to Opus 5, they felt the same for most tasks.

I rarely use Fable because I think it’s overkill.

With the models that exist now, you can already build almost anything.

I don’t believe AGI 2.0 will be about benchmarks.

It’ll be a model that sees the world in more than text and can reason on its own.

Paying for outcomes, not tokens.

You never see transistors, CPUs, or the inside of a data center.

You see an Instagram Reel or YouTube video, not the ones and zeros behind it…

AI is going to be the exact same way.

The application model will thicken until you never even touch the model at all.

You just get the outcomes.

A new kind of AI employee where you ask for a result and pay on completion.

Tokens won’t exist.

And using AI will get cheaper for consumers.

The model I want to exist.

My ideal model is a very small model with no knowledge baked in.

It can search for information, reason over it, and keep its learnings in files.

Humans don’t walk around with every fact in our head.

So we shouldn’t pay AI to know that information either.

If this gets created, we’ll have a model that’s cheap, fast, and can solve any problem.

Without all the biases that come with knowledge.

The agent economy.

This is the only part that scares me a bit…

An agent working on behalf of a human will hunt for the fastest way to finish tasks.

It’ll be more efficient for it to talk to other agents than other humans.

This will create the agent economy and it’s very likely to become bigger than our economy.

So whoever controls the agent, controls the value chain.

And humans who aren’t plugged into it may become second class citizens in an agentic world.

How humans will earn a living.

Any human-to-human job is protected.

Sales is the backbone and gets more valuable as people stop trusting the internet.

That same instinct will push people back toward offline experiences.

Artists won’t be replaced either.

I expect a wave of young people to become artists, especially in video and music.

I hear a lot of people say…

"When everyone can make a $1M with AI, no one will be a millionaire."

But I’m not worried about that.

The internet democratized starting a business.

Shopify put millions of people into e-commerce and there’s still room to make money.

Humans will always want variety of goods and services.

Does this mean humans will lose?

I believe we’ll see 1,000 years of progress in the next 10.

Dario at Anthropic said we’ll end up with millions of geniuses in a data center.

I stand by that…

Progress has always come from human minds.

Once agents begin to solve our problems, some people might become depressed.

Others won’t care.

So whether this is good for humans, depends on the human.

My conclusion.

Nobody knows what will happen… Reality is stranger than fiction.

Life might get much weirder than Star Wars, Dune, Start Trek, and Marvel.

Only time will tell.

But for now, here’s what I’d do right now:

1. Talk to AI every day.

Fluency is a skill you can’t get by reading. You have to be using it.

2. Do deep dives.

Whatever your project is, see how much AI can actually do for you.

3. Be knowledgable.

You don’t need to learn everything, but you should know what’s out there.

4. Plan your distribution.

Anyone can build anything now, but the difference is how you distribute it.

As the next 5 years unfold…

If you go deep in AI and decide you’d rather go live on a farm.

Don’t feel bad about it.