Hey Everyone - I’m back with an update. My technology and water fast was a bit of a failure. I actually didn’t miss the technology at all, I just stared at my wall and thought about food for 48 hours. Not the profound experience I was hoping for… would not recommend.

Still, life and AI will go on.

This week:

Let’s dive in.

There is a subtle force shaping which products win in an AI first world.

More and more people are not discovering tools through search results or blog posts. They are discovering them through AI assistants. The first framework an AI suggests. The database it scaffolds by default. The SDK it pulls into example code. These choices matter, because defaults compound.

That is what I mean by GEO. Generative Engine Optimization.

It is the practice of understanding what AI systems recommend, why they recommend it, and positioning yourself or your product accordingly. If an assistant reaches for your tool every time someone asks how to build something, you inherit distribution without ever running an ad.

You can already see early signs of this dynamic with Next.js.

Over the last few years, Next.js has climbed steadily in broad developer usage. Stack Overflow’s developer surveys show it rising from roughly 16.7 percent of respondents in 2023 to about 20.8 percent in 2025. Node.js has also rebounded strongly in that same window, jumping to nearly half of all respondents.

What makes this especially interesting is how those numbers shift among people who actively use AI. In the 2025 survey, both professionals and learners who reported using AI were more likely to report using Next.js than the overall population. That does not prove causation, but it is exactly the pattern you would expect if AI tools are shaping defaults for newcomers.

If you are learning today, you are probably not clicking through ten tutorials. You are asking an assistant to scaffold a project. The frameworks that appear in that first response get an enormous head start.

This is the strategic implication.

Products are now competing not just for human attention, but for model recommendations.

That changes how you evaluate technology bets. It also changes how founders should think about distribution. Documentation, example code, clean APIs, and opinionated starter templates suddenly matter even more, because they are what models learn from and reproduce.

It also opens up a new way to look ahead.

If you watch what AI systems recommend today, you get an early signal about where new users will flow tomorrow. That applies to frameworks, cloud providers, developer tools, marketing platforms, even creative software.

Which infrastructure keeps showing up in generated answers.
Which workflows get scaffolded automatically.
Which products become the assumed default.

Those are not random. They are paths being laid in advance.

So a useful habit in an AI shaped world is to regularly ask a simple question.

When someone asks an AI to solve this problem, what does it suggest.

If your product shows up, you are riding the current.

If it does not, you are swimming against it.

And if you are deciding where to invest your time, money, or learning, following those recommendations might be one of the fastest ways to see the future before it becomes obvious.

Lately, a lot of my conversations with friends who are building companies sound similar.

They are focused on product. Shipping features and trying to get something real into the world. At the same time, many of them are growing audiences in parallel. Posting online, writing newsletters, building personal brands. The problem is that the two efforts usually stay separate.

I understand that instinct because I have done exactly that before.

What I have become increasingly interested in is tying the two together. Building the product while growing the audience. Letting each side inform the other instead of treating distribution as something you bolt on later. I think it is even more important to try to do this for “free”.

I am still experimenting with this and helping a few people try it in practice, so I am not claiming this is solved. But the numbers make it hard to ignore.

If I had a year of runway, I would want to build the audience as quickly as possible and try to figure out how to build a product for them, rather than building a great product and then trying to figure out how to market it.

Here is how I think about it in concrete terms.

Let’s say your time is worth $100 / hr.

Option 1 - Purely organic. You spend 1hr writing a post. Maybe it brings in 5 subscribers. Maybe 10 on a good day. You have to create new organic content every week to stay relevant.

That is $10-20 per subscriber in opportunity cost.

Option 2 - Paid acquisition. You spend $100 on ads and pick up 100 subscribers in that same hour of setup and monitoring. That is $1 per sub plus maybe an hour of your time valued at $100.

You don’t necessarily have to keep creating new paid creative to get results. If you really only spend that 1 hour creating ads, you save a ton on opportunity cost.

Over Ten Weeks

Option 1: (10hrs x $100) / 100 subscribers = $10 per subscriber

Option 2: ((1hr x $100) + (10×$100(ad spend))) / 1000 subscribers = $1.10 per subscriber

Yes you spend more money, but you’re far more efficient with the spend.

Then there is the next layer.

Instead of just collecting emails, you offer something small but genuinely useful. A ten or fifteen dollar product. Give a discount code to anyone who joins the list.

Let’s say it costs you $1 to acquire each subscriber (removing the opportunity cost) and one out of every ten people buys your fifteen dollar offer. You spent $10 to get those ten people and you made $15. The list just paid for itself and you now know exactly who is willing to pull out a credit card.

You actually make $1.5 for every new subscriber. After your marketing costs of $1, you profit $0.50.

That $0.50 can go back into increased ad spend to grow your audience even more. You’ve created a flywheel that allows you to scale for “free”.

Over time you introduce something higher value. Maybe a two hundred dollar product. Maybe a cohort. Maybe a service. Then you have a real business.

Has anyone tried this?

You may not care at all, but I’m still doing my podcast the exponential athlete. I’ve been working on a new season, so this week I spent most of my time watching his interviews and a few documentaries on his background.

Things I Learned

This week’s survival skill is learning how to run small experiments with real money.

AI makes it easy to brainstorm, prototype, and spin up ideas quickly. What it does not remove is uncertainty about what people will actually pay for, show up for, or commit to. At some point, you have to leave the spreadsheet and test something in the real world.

I have become increasingly convinced that one of the best ways to prepare for an uncertain future is to get comfortable placing small bets. Spend fifty dollars on ads. Print a handful of flyers. Rent a small room for a meetup. Buy materials for a first batch of something and see if anyone wants it. The goal is not to make a huge return. The goal is to learn faster than everyone else.

Digital tests are obvious. Run a small ad campaign to a waitlist. Offer a low ticket product and track conversion. Try to get acquisition to break even. Watch what actually happens.

In person tests matter just as much.

Host a dinner and charge just enough to cover costs. Put a sign up at a coffee shop and see who scans the QR code. Bring a physical product to a local market and watch what people pick up and ask about. Offer a workshop at a coworking space. These kinds of experiments give you information you cannot get from dashboards alone.

What I like about this approach is that it changes how you think. You stop arguing with yourself in abstract terms. You stop waiting for perfect clarity. You stop polishing decks and start building feedback loops.

Even twenty or fifty dollars is enough to create signal. When someone hands over a credit card or Venmos you, you learn what actually matters to them. When they do not, you learn just as much.

This skill compounds over time. You get faster at designing tests. You get less emotionally attached to ideas. You build intuition around pricing, messaging, and demand. Most people never do this, which is why the people who do end up with such an advantage.

Bubble or not, AI or not, the ability to test ideas in the real world with real stakes does not go away.

Learn to run small experiments with real money.