AI helped us build it. We were still guessing who wanted it.

We've been exploring a game for Club Spite, with AI helping us get playable versions together. Characters moving, enemies spawning, weapons firing. If something looked wrong, we could change it and try again. We even tried a different engine.

We'd also discussed where the game could live and how it might make money. Steam, browsers, ads. There was plenty to work on. We were still trying to pin down why someone would choose our game over the ones they already enjoy.

The enemies were showing up reliably. The players were still an assumption.

We've looked at other games and read what people like about them. That's useful research. It establishes that people enjoy that kind of game. It doesn't give us an audience of our own, and bolting another mechanic onto it won't necessarily change that.

A weapon can look better and behave more convincingly without making a game worth playing. That's an annoying possibility to consider when you've just finished fixing the weapon.

I caught myself wondering whether we were developing a good idea or keeping one going because making another version had become so easy. I think that's worth asking before we spend the next few weeks making the menu nicer.

It used to be harder to ignore the commitment

I've spent more than 30 years across development, DevOps, security and programming. Building an app meant a serious investment in your time, someone else's time, or both. You had a reason to think carefully before taking it on.

That didn't stop people building things nobody wanted. Expensive development wasn't a quality-control department. It also kept people with good ideas out because they couldn't code, couldn't pay, or couldn't spare the months. I wouldn't bring that barrier back.

Being able to try something without betting half your year on it is a real improvement. Make a game because you want to make a game. Build a ridiculous little site because it amuses you. Enjoying the work is a perfectly good reason to do it.

Once I'm planning a business around it, though, I'd like some evidence that the enthusiasm extends beyond my own desk.

That's where AI makes it easy to dodge an awkward question. I can ask for a feature and get something concrete to inspect. I can fix a bug and watch it disappear. Working out whether people want the product is less cooperative. They might already have something they prefer. They might like the idea and still never use it.

There is an appealing alternative: go back and improve the product again. Something will have changed by the end of the session, and I get to feel productive.

If I keep doing that and call it market research, I'm bullshitting myself. It's development. Useful development, possibly, but it hasn't answered the question I was avoiding.

Everyone else gets more software to sort through

RevenueCat's 2026 report includes Appfigures data showing monthly subscription-app launches rising from about 2,000 in January 2022 to more than 14,700 in January 2026. The chart doesn't identify which apps were AI-built or how much customer research went into them. It does show a substantial increase in things competing to be installed. RevenueCat, State of Subscription Apps 2026

That matters at the other end of the download. Making an app cheaper for me to produce doesn't make it effortless for you to try. Depending on the product, you might have to create an account, move some data and learn enough to find out whether it does the job. A subscription gives you something else to remember to cancel if it doesn't.

Those are reasonable things to do for a product you want. They're a lot to keep doing for products whose creators are still figuring that part out.

A May 2026 preprint examined more than 160,000 Product Hunt launches. Its authors found increased solo entry after ChatGPT's release, without a corresponding increase in solo representation among the highest-ranked launches. Those rankings can't tell us whether a product paid its bills or kept its users. They do give us a reason to separate the number of things launched from how well they do on the platform afterward. Kim, Kang and Song, 2026

Neither source tells us how many builders skipped market research. I'm not going to manufacture that conclusion because it would make a convenient headline. Some of those new products could be excellent. A smaller market can also be worth serving now that the work costs less.

What I'm questioning is the leap from being able to ship something to assuming there must be people waiting for it. For anything I hope to launch as a business, I'd want to find out what would make someone use it before planning the first year of updates.

Give the prototype something to find out

Market research doesn't have to mean hiring a consultancy or refusing to write a line of code until a spreadsheet grants permission. A cheap prototype can be part of the research. Someone trying it can tell you more than another week spent arguing about its potential.

For an app, I'd want to know what people currently use and what would make changing worth the hassle. Complaints are worth investigating, but someone being fed up with a tool doesn't mean they'll install mine. The thing they hate might still do the one job they can't give up.

With our game, I'd want people who play this kind of thing to try it without me explaining why every mechanic is supposed to be interesting. I'd watch where they get bored and whether they choose another run. If they leave, I'd want to understand that before adding more levels for them not to play.

AI can help find competitors, organize real feedback and build the experiment. I'd happily use it for all of that. Having it predict how enthusiastic my future customers will be leaves the customers out of the exercise, which is a fairly significant omission.

I don't want less creativity or fewer people getting a chance to make something. I do want to notice when an enjoyable experiment has become a business plan whose evidence consists mostly of the fact that the software runs.

For the game, I can already see plenty more things we could build. Before another batch of features, I'd like to see someone with no involvement in making it choose to come back. That would give the next round of work a much better reason than the fact that I can ask for it.