Two years into the AI coding assistant boom, a reasonable person might ask: where are all the useful apps?
Tools like Cursor, Claude Code, and GitHub Copilot have made it genuinely faster to write code. Solo developers can now build in a weekend what used to take weeks. The barrier to shipping software has never been lower. And yet, browse any app directory or product launch site, and you'll find the same pattern repeating: AI wrappers, chatbot skins, and "AI-powered" versions of things that already worked fine.
The $50 App Problem
The pre-AI indie software scene already had a strong track record. Developers routinely built focused, useful tools on tiny budgets. One well-known example: a developer who built a COVID vaccine appointment finder for about $50 in infrastructure costs. Simple, specific, immediately useful.
That kind of project didn't need AI coding assistance to exist. It needed someone who noticed a real problem and knew enough to solve it. The bottleneck was never "writing code is too slow." It was finding problems worth solving and doing the unglamorous work of making software that real people actually use.
More Code, Same Number of Good Ideas
AI coding tools solve the production bottleneck. They make it faster to go from idea to deployed app. But they don't solve the ideation bottleneck, and they definitely don't solve the distribution bottleneck.
What we're seeing instead is predictable: lower friction to build means more things get built, but the ratio of useful-to-useless stays roughly the same. Maybe it even gets worse, because the people who previously would have given up halfway through a mediocre idea now finish it.
The directories are flooded with AI-generated SaaS products that solve problems nobody has. "AI meal planner" number 47. Another "AI resume builder." A ChatGPT wrapper with a slightly different UI and a $20/month price tag.
What Would Actually Change Things
The interesting question isn't whether AI coding tools work - they clearly do. It's whether making software easier to build changes what gets built.
So far, the answer seems to be: not much. The developers who were already building useful things are building them faster. The developers who were building derivative things are building derivative things faster. And a new wave of non-developers are building their first apps, which are mostly clones of existing apps.
There are exceptions. AI coding tools have genuinely enabled some non-programmers to build internal tools and automations they couldn't have built before. That's real value, even if it doesn't show up on product launch sites. A small business owner who uses Bolt or Cursor to build a custom inventory tracker isn't posting about it online - they're just using it.
The most honest assessment might be this: AI coding assistants are a productivity multiplier, not a creativity multiplier. They make good developers more productive and make it possible for non-developers to build simple tools. But they haven't changed the fundamental economics of what makes software worth using. You still need a real problem, a clear solution, and the patience to make it work well. No amount of AI-generated code changes that.