I vibe coded a feed reading web app. It was enlightening and uncomfortable

The experiment

It has been reported that Thomas Claburn bought a $20/month Claude subscription and used it to "vibe code" a web app for monitoring news feeds — an exercise he describes as both enlightening and uncomfortable. The result, he says, wasn't perfect or artful, but it worked: code good enough to ship. Short answer: AI-assisted coding now moves projects from idea to prototype faster than many expected. The kicker? It felt oddly transactional, like hiring a competent contractor who neither cares nor judges.

The uncomfortable truth

Claburn walks through a broader arc: 2019’s quirky AI outputs, 2022’s passable code and legal headaches, and a 2025 moment when Andrej Karpathy allegedly coined "vibe coding" for machine‑generated, pragmatic code. By late 2025 — with model releases such as Anthropic's Opus 4.5 and OpenAI's Codex 5.2 — the line between "vibe coding" and ordinary coding blurred. Developers noticed, GitHub commits ballooned, and a flood of testimonials followed: people building useful stuff with models that are competent but not elegant. Is craftsmanship being sidelined? Maybe. But punk and rap once made similar people wince — tastes and tools evolve.

What it means

Claburn doesn’t romanticize the tech. It has been reported that he believes the risk isn’t AI itself but those who deploy it without accountability, and he argues that meaningful checks will require politics as much as policy — "stopping AI in its present form begins at the ballot box," he writes. So yes: vibe coding works, and that’s both a relief and a wake‑up call. We get faster prototyping. We also get ethical and legal questions landing harder and faster than many devs expected. Careful craft still matters — but for how long, and who gets to decide?

Sources: The Register