Bubble-Burst.

You are making two big mistakes.

1. You think AI is a bubble.
2. You overestimate how hard it is to master AI.

I am fixing these two in one article.

Before diving in, take a second to share this with your colleagues to finally answer: Is AI a bubble?

1. It is not a bubble.

Everything feels fast. Too fast. So it must be a bubble.

First, we did not wait for AI to happen to go too fast.

For centuries, a person could live 250 years after their ancestor and recognize every part of their daily life. Today, just 25 years can make the world feel unrecognizable. We do not live in the 2000s anymore.

The 2000s… remind me of another bubble.

See, I am the AI guy for the people around me. So when they meet me, they feel the urge to warn me, “Ruben, remember dotcom! it’s a bubble!! be careful.”

It’s been 4 years, and it never fails to amaze me.

People spend a lifetime preparing for a bubble. But either you're right on time to call out the bubble, or you’re falling behind.

This is how I see them:

So in February, I wrote: “Thinking AI might be a bubble is the most dangerous belief you can hold in 2026.”

I like to think of bubble/not bubble through 4 lenses:

1. Adoption: Are more people using AI, more often?
2. Monetization: Are people spending more money on AI?
3. Intelligence: Is AI becoming more capable?
4. Affordability: Is the same capability becoming cheaper?

We went from 0 to +50% of businesses using AI in 4 years.

And AI spending quadrupled this year (from Feb. 25 to Feb. 26).

Impressive, but this is nothing compared to intelligence & affordability.

Better & 79x cheaper.

AI became better, and 79× cheaper in just one year:

And when you compare the two models, it is fascinating.

First, GLM is clearly better than Opus 4.1:

Now, if you want to get technical, we can compare the cost per task to complete each benchmark.

In other words:

1. AI went from 0% to 50% adoption in 4 years.
2. AI spending quadrupled in 12 months (Feb-Feb).
3. AI got a bit better while being 79x cheaper in one year.

Now comes your objection: “But I pay for Claude, and it keeps getting more expensive!!” Yes. Because you use the latest Fable-5.1, a much smarter model.

And you use it a lot, a whole lot more than last year.

But eventually, you won’t need PhD-level models. Or Einstein-AI.

The same level of intelligence will get you very far.

Look at accounting, for example:

Mercor hired 12 accountants to complete 4 realistic accounting tasks, then had AI models from the past few years complete the same work.

As of May 2025, accountants outperformed AI.

Now they lose out to today's AI models, which are near-perfect.

So take a second to understand.

• May 2025: the AI models are close to good enough.
• May 2026: AI is way better than any human at the task.
• May 2027: the same AI is now 79x cheaper than last year.
• May 2028: the same AI is now 79x79= 6241x cheaper than 2026.
• May 2029: it’s now 493,000x cheaper at the same task than 2026.
• May 2030: an accounting task that used to cost AI a million dollars in 2026 is now… 3 cents.

Is your accountant getting 79x cheaper, every single year?

This is not normal.

No other technology got so cheap so fast.

Experts call LLMs (the technology behind today’s AI) too cheap to meter.

Imagine my son telling me that he’s pursuing accounting.

I would say, “Listen son, accounting is dead. This is no longer a skill that we deem valuable.” We used to have people carrying luggage around. I hope my son won't be one of those.

I want him to be useful today and for the foreseeable future.

Since I don’t have a kid yet, I want us to be useful today & tomorrow.

And if you don’t know where to start, I have good news for you.

2. You only need 23 minutes.

If you were a complete rookie with AI, or you’re confused about how to get started, how fast can I teach you most of it?

It took me weeks to answer this. And about a dozen cohorts.

I could tell you how hard AI is. Very hard.

I could tell you that you are behind. Very behind.

I could tell you that AI is full of secrets & mysteries. Only I have the answers.

I could tell you I recorded hundreds of hours to decipher (I would use the word ‘decipher’ to sound intellectual) the wonders of AI.

But I don’t.

I respect your time. You are a busy professional.

AI is getting easier with time. Not harder.

You just need to know how to start properly. Like a bike with wheels.

It takes 23 minutes to bike by yourself.

1. 23 minutes to go from an absolute rookie, a zero, to 80% there. This is my Monday workshop called “AI 101”. We go through the entire UI of Claude, so you know how to use every tab and button.
2. 23 minutes to go from 80% to 95% there. This is our “AI 201” class to master Claude Code and how to best use it (with GitHub, Vercel, and other apps connected).

I could charge $200 for each class. But it does not feel right. In 6 months, you will come back and will have to pay for a new class again to start over.

So instead, we have these two live sessions every single week, and they are recorded. All of this for the same $200 a year subscription.

Save your seat for next week live cohort beyond this paywall:

PS: I connect on Linkedin with everyone joining the live! I want to stay close to my audience. I will see you on the other side: