Can AI build a real web application from a single prompt? I recently experimented with this idea by developing a personal librarian app using tools like ChatGPT, Copilot, and Codex. In less than 30 hours, the project went from an initial prompt to a working application—and along the way it revealed where AI dramatically accelerates development and where engineers still need to take the wheel.
Google’s generous free-tier limits once made Gemini an ideal platform for experimenting with LLM-powered applications. When those limits quietly changed, a YouTube recommendation system that had worked reliably for months suddenly stopped functioning. This post recounts that experience and explores what it reveals about the hidden risks of building on third-party AI platforms.
I am not a web designer by training, yet web design has been a recurring part of my work for nearly two decades. Building single page applications means dealing with HTML, CSS, and layout decisions whether one enjoys it or not. For a long time, I approached web design pragmatically, learning just enough to make things work, while never feeling fully in control of the result. Recently, that changed.