Make or AI

Make or Buy was yesterday.
Today the question is: Make or AI.

And that is not a semantic detail. It is a paradigm shift.

In the past, the decision was obvious: we do not build a cookie consent banner ourselves.
Not with two people.
Not alongside the day job.
Not without specialized expertise.

That was infrastructure. Complex. Legally sensitive. Full of edge cases.

So: Buy. Drop in the script. Pay the monthly fee. Done.

The same applied to explanatory product flows, micro features, or content modules. Anything important but not directly differentiating was outsourced.

Not because we could not build it.
But because it was not realistic.

With two people, something like that simply was not plannable.

Saturday, waffles, and consent logic

Today I build a cookie consent banner including the full underlying technology myself.

Not as a quarterly initiative.
Not as a roadmap epic.
But on a Saturday between grocery shopping and making waffles with the kids.

With Claude Code.

UI. Consent states. Storage. Versioning. Integration into tracking flows.

What used to feel like a multi day mini project is now an iterative dialogue with AI. Prompt. Review. Adjust. Deploy.

The boundary did not shift.
It no longer exists in the same way.

Two evenings, fifty articles

This week, over two evenings after the kids were in bed, I built a complete help center for Joinride.

Structure. Navigation. Product integration.

And around 50 articles.

SEO optimized. Agent optimized.
Tailored precisely to our product, our voice and tone, our glossary.

The result is live at
https://joinride.cc/support

A few years ago, that would have been a dedicated project. With a writer. With an SEO brief. With backlog grooming. With weeks of coordination.

Today it is two focused evenings.

AI changes how we think

We used to think in sprints.

Do we have capacity?
Is this an MVP?
What can we cut to make it fit into two weeks?

Now I think differently.

Cookie consent? I will ship it by tonight.
Help center? I will build it this week after work.

AI does not move the line between Make and Buy.
It shatters it.

For a two person project like Joinride, that means infrastructure is no longer automatically a Buy case.

There is effectively a virtual team sitting next to us.
Boilerplate gets written.
Edge cases are considered.
Content gets structured and optimized.

And we keep full control.

How big does a topic need to be?

Of course, not everything should be built in house.
Highly regulated areas, specialized infrastructure, deep tech problems will remain Buy candidates.

But the threshold has fundamentally changed.

The more interesting question today is no longer: Make or Buy?

It is:

How big does a topic actually need to be before I seriously consider buying it?

If two people can build infrastructure, content systems, and knowledge bases at product quality between everyday life, family, and work, then it is clear:

This is not a tooling upgrade.
It is a structural shift.