Osegi
AI·2 min read

Vibecoding lets users build the apps developers never made

AI chatbots now let anyone build tailor-made apps by typing a request. It shifts power from app developers to AI model providers, with trade-offs in support and security.

A person describing a custom app to an AI chatbot on a laptop, with a smartphone showing a tailor-made interface beside it.
Image: xataka.com

For years, an app did only what its developer decided. To compete, developers kept adding features until many apps became bloated, and users had little recourse beyond the suggestion box. AI-assisted "vibecoding" is now changing that balance.

The logic is simple: an app usually takes off by solving one specific need, such as reminding you to drink water, switching on a flashlight or sharing photos. Once dozens of rivals do the same, the only way to stand out is to add more. And no developer can tailor software to every individual user. With a large language model acting as a bespoke tailor, users can now request the exact button they want.

Not new, but far easier

The author of the analysis says he has already built several apps this way: one that controls home automation without data leaving his network, one that removes bloatware from Android phones, and a family task app backed by a database on his own server.

Custom automation is almost as old as software itself, from Excel macros to IFTTT, Tasker on Android and Shortcuts on iOS. What changes is that nothing needs to be learned first: you write what you want in plain language.

Power moves to the model makers

The developer's "dictatorship" fades, but freedom is only partial. Power now sits with the providers of AI models, namely OpenAI, Anthropic, Google, xAI and others. What looks like a commodity can become a dependency, and dependencies eventually get billed. A good vibecoded app calls for a good model, so the subscription does not vanish. It moves from the developer's app to the AI plan.

There are practical limits too. A personal app may work without its owner fully understanding why. There is no support, bug fixing depends on the model, and nothing guarantees it won't break something sensitive.

The author does not expect vibecoding to replace developers, just as image and text generation haven't replaced designers and writers. It gives you an app that works for you, not one that scales to thousands of users, stays secure or gets maintained over time. In the short term, mainstream apps will keep covering what most people need, while vibecoding fills the gaps no developer can reach.

Apps worth a look

Tags:vibecodingaiappsopenaianthropiccustom software

More from this day

See all →

More in: AI

See all →