The Kind Truth Behind Vibe Coding.
The internet really romanticized “vibe coding”.
So here’s the honest version from someone deep inside it:
I’ve consumed more than 7.3M AI tokens building my SaaS… and the MVP still isn’t finished.
Not because AI is bad.
Not because the idea is impossible.
But because once you stop building simple landing pages and start building an actual builder/editor with real logic, infrastructure, persistence systems, state management, autosave pipelines, databases, sync flows, caching, architecture decisions, and scaling concerns…
everything changes.
The UI part is easy.
The system behind the UI is the real monster.
And the craziest part with AI:
every time you solve one problem, you suddenly discover three more behind it.
You fix persistence…
then uncover race conditions.
You fix state sync…
then discover architectural flaws.
You optimize performance…
then break another flow somewhere else.
AI can generate code fast.
But maintaining coherence across a growing codebase is a completely different game.
Even when I check /status during my own sessions, it reminds me of the scale of what I’m doing:
“Your longest session is ~30x longer than a World Cup soccer match.”
At that point, it stops being “vibes” and starts being long-form engineering endurance.
Still one of the craziest learning experiences I’ve ever had.
Would genuinely love to hear experiences from other people building complex products with AI — especially the parts nobody talks about publicly.
