Ideas and taste are the job now

I'm going to be honest: the short version of this post sounds like every LinkedIn post you scrolled past this year. "Taste is the new skill." "AI changes everything." I've rolled my eyes at those too.

So this is the long version, with drawings and receipts. By the end you should know what I mean by ideas and taste, why I think they matter more every month, and what I actually do to get better at both.

What building used to take

A few years ago, an app idea needed a team before it was anything: someone to design it, someone for iOS, someone for Android, someone for the backend, someone to test it. That's months and a budget before you find out whether the idea was any good.

Now one person with a laptop and a good coding agent can get from idea to App Store. Not a toy. A real app with accounts, payments and a privacy policy.

The bottleneck moved

Building didn't become free. But it shrank a lot, and when one step shrinks, the steps around it become the slow part.

The two ends are deciding what's worth building (the idea) and deciding when it's actually good (taste). Those are the jobs AI can't take off your plate, because they're the ones where you are the user.

What I mean by "idea"

Not a billion-dollar vision. Most of my good ideas started as small annoyances: a warranty I forgot to claim (Receipt Snap), an invoice I didn't want to type (Invoicely), a lock screen that looked like everyone else's (Lock Card).

An idea is just noticing a problem clearly enough to describe the fix in one sentence. These days, that one sentence is most of the spec.

What I mean by "taste"

Taste isn't liking nice things. It's a filter.

Every version will compile. Most will be fine. Taste is knowing which one is right and being able to say what's wrong with the rest: too many colours, a button that jumps on every tap, an empty state that just says "nothing here".

The better the tools get, the more versions you get, and the more the filter matters.

I keep my filter written down as rules, so I can hand it to an agent too: one press, one spring and why some apps feel expensive.

Receipts

This isn't a theory for me. These are my own apps, live on the App Store, each one designed and built alone.

AI did a big share of the typing. It did none of the deciding: which ideas were worth building at all, what to cut from each one, and when a screen was finally done. That part was all ideas and taste.

The gap

This is the part I'm least comfortable saying, because it sounds like a threat. It isn't. It's what I see around me.

People who use AI every day compound. Each project teaches them what the tools are good at, which makes the next one faster, which leaves more time for ideas and taste. People who ignore it aren't going backwards, but the line next to them keeps bending upward. The space between those lines is the gap, and I think it keeps growing.

How to get better at both

• Write your annoyances down. Ideas come from noticing. A note on your phone is enough.
• Ship small things. A tiny app that's finished teaches more taste than a big one that isn't.
• Use great apps closely. Notice what they leave out, not just what they add.
• Write your taste down as rules. If you can't explain why a version is wrong, you can't teach an agent either.
• Use AI daily, for real work. Not to watch demos. To learn where it's good and where you still have to step in.

tl;dr

• AI made building cheap. It didn't make knowing what to build cheap.
• An idea is a problem noticed clearly. Taste is the filter for what's good.
• One person with ideas, taste and AI can now ship what used to need a team. I've done it nine times.
• The people learning this now will be hard to catch.

More on the day-to-day: how I build with AI. And if you have an idea and want help bringing it to life, that's what I do.