Ok, I’m definitely late to the trend, but I finally dipped my toes into the whole vibe coding hype… :sweat_smile:

I gave myself just a few hours (while half-distracted with other stuff) to prompt my way through building FrenchPing, a tiny game to practice French, wrapped in the Teenieping universe. I built it entirely with GPT-5 and Co-Pilot in Agent mode, without typing a single line of code myself (that was the challenge).

At first, it’s honestly quite impressive. Watching it spin up a UI feels great, especially if like me UI design isn’t really your strong suit. What surprised me, though, was that instead of defaulting to the usual suspects (Next.js, React/Vite, etc.), it just went old school with plain HTML, CSS, and JavaScript. Not what I expected at all :thinking: :shrug:

The real headache came with TTS. Pronunciation for single words was pretty rough without context. For instance, it kept saying “pain” the English way instead of the French pain (“bread”). I tried nudging it towards ElevenLabs with a decent French TTS model, but that only seemed to make things messier. The agent started throwing in weird fallbacks and pronunciation hacks everywhere. Long story short: the TTS is not in a great place right now.

As for the code… well, let’s just say the more you prompt, the messier it gets. Even when you ask for refactors, it rarely comes out cleaner or more optimized. Plus, the agent flat-out avoided any server-side rendering, and constantly drop the API keys directly into the frontend code. Sure, they were free-tier API keys, but imagine a true non-coder vibe-coding like this with a paid API… yikes.

Anyway, it was a fun little Friday experiment.

See ya, in a future post 👋