The plan

I told myself this would take 30 minutes. Add authentication, hook up a welcome email, maybe set up a simple onboarding flow so when someone signs up they actually know what to do next. Nothing crazy, just enough to move the product forward.

My kid was at daycare until 1. My wife was watching trash TV. I had a clean window and figured I’d knock this out before joining her on the couch like a civilized human. This is the kind of work that should be straightforward now, right? AI writes the code, tools are better than ever, everyone keeps saying how easy it is to build. So I figured… let’s test that.

The build

I opened Cursor, pointed it at my codebase, and just started going. No overthinking, no architecture doc, no “let me plan this out for an hour.” Just straight into it. Paste the docs, follow the steps, swap out my janky hand-coded login component for something more legit.

And honestly… it felt smooth as hell at first. Stuff was happening fast. Components updating, auth flows wiring up, buttons appearing. Things were working. I could click through it, register, log in. It looked like a real product.

And that’s the dangerous part. When things are moving fast, your brain assumes you’re doing the right thing. You’re not thinking, you’re reacting. You’re stacking decisions without really understanding them. AI is incredible at helping you do that. It removes friction, which sounds great until you realize friction was the thing forcing you to think in the first place.

The oh shit moment

At some point I paused. Not because something broke, nothing broke, which made it worse. I paused because something felt off. So I stepped back and looked at what I had actually just built.

I had layered a full authentication system on top of another authentication system.

I was already using Supabase for auth and user management, and without really thinking about it, I started integrating another tool that was also trying to handle identity. Two sources of truth, two systems managing users, two different ways of thinking about sessions.

Which, if I kept going, would’ve turned into a complete shitshow. Conflicting user states, broken redirects, edge cases that don’t show up until production and then you’re sitting there wondering what the hell is going on.

And the worst part is nothing in the moment told me it was wrong. Everything looked fine. That’s the trap. AI helps you build things that look correct. It doesn’t tell you if the thinking behind it is garbage.

The pivot

So I stopped, undid everything, ripped out the second system completely, and went back to Supabase as the single source of truth. It felt like going backwards, but it was actually the first correct decision I made the entire session.

Once I made that call, everything got simpler. Auth worked, sessions made sense, the login flow was clean, and I could actually start thinking about onboarding instead of just “does this technically function.”

And here’s the funny part. From the outside, it looks like I just built auth in 30 minutes, which is technically true. But what actually happened in those 30 minutes was make the wrong decision, go pretty far down that path, catch it before it becomes a real problem, rip it out, and rebuild it properly.

That’s the real work.

The real shift

This is the gap right now. Everyone can build, but very few people know what the hell they’re building. Especially designers. We’re used to thinking in flows, screens, and interactions, which is useful, but once you start building you’re forced into a different layer of thinking.

Where does this live? What’s the source of truth? How does this connect to onboarding? What triggers retention? How does this eventually make money?

Because auth isn’t just “login works.” It’s the entry point into your entire system. It’s what triggers your welcome emails, kicks off onboarding, and determines how you identify and track a user across your product. If you get that wrong, everything built on top of it is shaky.

AI doesn’t solve that. It helps you move fast. It doesn’t help you think better. That’s still on you.

Why this matters (especially if you’re a designer)

If you’re trying to get into the entrepreneurial side of things, this is the shift. It’s not about becoming an amazing engineer overnight. It’s about understanding systems and how a feature connects to acquisition, onboarding, retention, and ultimately revenue.

In the same session, right after fixing auth, I was already thinking about what happens when someone signs up, what email they get immediately, how I guide them over the next 7 days, and how I get them to actually use the product instead of just bouncing.

That’s where the leverage is. Not just building the thing, but designing how the thing actually works as part of a system.

Watch it

I recorded the entire session. No polish, no pretending I knew what I was doing, just the real process including the part where I almost built something completely wrong and had to catch myself mid-way.

If you’re a designer trying to get into building, or already messing around with AI tools and feeling like something is off but you can’t quite explain it, this will probably feel very familiar.

Watch it here:

​

​
Talk soon,