Figma Make slowed me down more than it helped

Figma Make pissed me off this week. Like actually pissed me off.

I was trying to build something real. Not a hero section. Not a fake UI. Real product shit. Upload, processing, validation, graph, review. Something with actual moving parts.

And it felt wrong immediately. Not visually. Physically.

Like dragging your hand across rough paper. That dry resistance. That slight catch on your skin. You feel it before you even think about it.

Every interaction had that same friction. Nothing was clean. Nothing flowed. You could feel the tool pushing back.

And every small change cost me. Credits gone. Again. And again. I burned through two full rounds and had to get three top ups just to keep going.

And I am not bad at this. I am direct as hell with prompts. Still didn’t matter.

Stuff would hang mid generation. You just sit there staring at it. Waiting. Watching it eat credits.

And when it finally gives you something back, it’s off. Not completely wrong. Just wrong enough.

Half right. Half useless. Close enough to trick you. Not good enough to use.

Credits gone.

At some point I just sat there like… what the fuck am I doing.

It stopped feeling like design. It felt like being stuck in something you don’t control.

Pull the lever. Hope it works. It doesn’t. Try again.

Same motion. Same result. Slightly different failure each time.

You start second guessing yourself. Maybe it’s the prompt. Maybe it’s you. It’s not. It’s the system.

That was the moment I stopped. And that’s when I realized the real problem.

Because up until that point, I was still trying to make it work. Tweaking prompts. Trying different angles. Hoping the next iteration would click.

It never did.

This wasn’t a tool problem. It was a design problem

I wasn’t building something simple. This was a system. Things connect. State moves. Users move through it. And you can fake that in Figma. You can make it look right. But once it’s real, it breaks.

Wireframes hide that. Mocks hide it even better. And yeah, I’ve been part of that.

Because you can make anything feel smooth in Figma. Everything snaps. Everything aligns. Everything looks finished.

Everything feels right.

Until it has to actually work.

Until it isn’t.

So the tool didn’t fail me. My process did. That’s what forced the switch.

And once you see that, you can’t unsee it. You realize how much you’ve been relying on something that lets you get away with it.

Talking it out showed me more than designing ever did

I opened ChatGPT voice and just talked. No prompts. Just explaining the product like I would in a meeting. Upload, processing, validation. The admin uploads a batch of PDFs. Nodes validate logic. Users think in files, not forms.

It was messy. But it was real.

Then I turned that into a PRD. Saved it as markdown and dropped it into the project. That file became the source of truth. And everything changed after that.

Because now I had something concrete. Not a feeling. Not a loose idea. Something I could actually build against.

Building in code forced the truth out

I opened Cursor and started building. One piece at a time. Upload, processing, graph, inspector, review. Build it, use it, break it, fix it, repeat.

And it felt way better than the other loop. Because now I couldn’t hide.

There was no illusion anymore. Either it worked or it didn’t.

No smoothing it over. No pretending it works. If it breaks, it breaks right in front of you.

And that’s when it broke.

Everything broke once it was real

Too many nodes. Wrong mental model. Numbers didn’t match. Flow didn’t make sense. It felt off.

Not visually. Fundamentally.

Like wearing something that looks fine in the mirror but feels wrong the second you move. You can’t ignore it once you feel it.

These weren’t design tweaks. These were product decisions. And I only saw them because it was real. That’s the part people miss.

You don’t find these problems by thinking harder. You find them by running into them.

Fixing it meant changing the system

Not colors. Not spacing. The system. Nodes became files. The summary step got killed. The inspector became contextual. Layout mattered more than styling. I switched to React Flow. Added guided review.

All obvious once it was real. None of it obvious in Figma.

That’s when it clicked.

The goal isn’t to make something look right.

It’s to make something that holds up when it’s real.

Spec driven development is just a better loop for this

Not always. But for real products.

The loop is simple. Talk it out. Write the PRD. Save it as markdown. Build in small pieces. Use the product. Fix what breaks. Repeat.

No fake certainty. No pretending. Just truth.

And truth is uncomfortable at first. But it’s way faster.

What this looked like in reality

In Figma Make it was burn credits, wait, hope, repeat. You feel stuck in it.

In ChatGPT voice it was clarity, structure, PRD.

In Cursor it was build, break, fix, repeat. You feel back in control.

And honestly, I like this way more. Flat cost. No credit anxiety. No guessing. Just me and the code where I should have been. Real libraries. Real components. Real constraints. That’s the difference.

It feels like working with the product instead of negotiating with a tool.

This changed how I design

I’m not starting with wireframes anymore. Not for this kind of work. Too easy to lie.

If it matters, build it. If it breaks, fix it.

AI didn’t replace anything. It just made this faster. Taste still matters. Decisions still matter.

Anyone can generate screens. Not everyone can build something that works. That gap is real. And it’s getting bigger.

And the people who figure this out early are going to move a lot faster than everyone else.

If you want to see what this actually looks like when you go from nothing to a real product, we broke down the full experience of building and launching a SaaS product from scratch in this week’s episode.​

Different context, same realization. Once something is real, everything changes.

Till next week,