The biggest bottleneck isn't coding anymore.

IL — Mon, 06 Jul 2026 13:28:02 +0000

A few years ago, building software was the hard part.

You needed months.

Sometimes years.

A team.

Infrastructure.

A budget.

Today?

One developer with Cursor, Claude Code or ChatGPT can build in a weekend what used to take weeks.

That's amazing.

But it also revealed something I wasn't expecting.

Coding was never the real bottleneck.

It was just the most visible one.

The real bottleneck starts after you ship.

Shipping isn't the finish line anymore.

I've seen people spend months building AI products.

Beautiful landing pages.

Great UX.

Solid code.

Then they launch.

Nothing happens.

Not because the product is bad.

Because nobody knows it exists.

It's a strange feeling.

You finally finish building...

...and realize you now have a completely different job.

Distribution became the hard part.

Getting the first customer.

Then the tenth.

Then the hundredth.

That's the work AI didn't automate.

SEO.

Writing.

Communities.

Partnerships.

Word of mouth.

Cold emails.

Videos.

Podcasts.

Answering questions on Reddit.

Showing up every day.

None of those things got dramatically easier.

AI made one thing incredibly cheap.

Building.

That's fantastic.

But something interesting happened.

When building becomes cheaper...

building stops being the competitive advantage.

Think about it.

If ten companies can build the same idea in a week...

why would yours win?

Probably not because your button has a nicer border radius.

Or because your backend is written in a newer framework.

People don't discover products because the architecture is elegant.

They discover them because someone talked about them.

Found them on Google.

Saw them on Reddit.

Watched a video.

Read an article.

Heard about them from another customer.

Distribution compounds.

Code doesn't.

I think we're entering a different era.

Developers used to ask:

"Which framework should I learn next?"

I'm starting to think a better question is:

"How will people find what I build?"

That's a much harder question.

And a much more valuable one.

This changes what being a founder means.

Years ago, being technical gave you a huge advantage.

Today it gives you speed.

Those aren't the same thing.

Speed helps you build faster.

Distribution decides whether anyone cares.

That's why I keep seeing incredibly talented developers struggle to get customers.

They're competing in a game that changed without realizing it.

The bottleneck moved.

The weird part

I don't think AI made startups easier.

I think it changed what "hard" means.

Writing code is getting cheaper every month.

Attention isn't.

Trust isn't.

Distribution isn't.

Maybe that's why the founders who stand out over the next few years won't be the ones writing the most code.

They'll be the ones who learn how to get their products in front of the right people.

Because AI didn't eliminate the hardest part.

It simply moved it somewhere else.

I built hundreds of AI-generated websites. Here's what surprised me.

IL — Thu, 02 Jul 2026 19:06:14 +0000

When I started building an AI website generator, I thought the biggest challenge would be the prompts.

I was wrong.

The prompts were actually the easy part.

The surprising part was everything around them.

For example, I expected AI to struggle with layouts.

It didn't.

After a few iterations, the layouts were good enough.

What consistently failed was understanding the business.

A plumber doesn't need the same homepage as a dentist.

A coffee shop shouldn't look like a law firm.

That sounds obvious.

It wasn't obvious to the AI.

So I stopped spending time improving the HTML.

I started spending time improving the context.

Instead of asking the model to "build a website", I gave it information about the business.

Its industry.

Its services.

Its goals.

Its customers.

The output changed dramatically.

That taught me something interesting.

Website generation isn't really a design problem.

It's a context problem.

The better the context, the less the model has to guess.

Then another problem appeared.

The generated website looked good.

But business owners still wanted to change things.

Opening hours.

Services.

Phone numbers.

Photos.

That's when I realized something else.

Generating a website is a one-time event.

Maintaining it is the real product.

Most website builders optimize for the first five minutes.

Users spend the next five years editing that website.

That's the experience that actually matters.

AI made creating websites incredibly easy.

Now the challenge is making them stay useful.

And that's a much more interesting problem to solve.