Mateusz Konik

I'm a veteran and solo founder with no formal coding background. Spent the last few weeks directing Replit's AI agent (with Claude as a strategic/QA partner) to build a real production app, entirely from my phone.

The part that surprised me: the "vibe" is the easy 20%. The other 80% was the same discipline any build needs catching bugs by hand (missing auto-capitalization, a timezone off-by-one, a file upload limit rejecting real phone photos), refusing to accept "it's fixed" without proof, running full walkthroughs section by section, chasing root causes instead of patching symptoms.

That's made me rethink what "vibe coding" even means. It's not skipping the engineering discipline it's relocating it. Instead of writing code, you're writing precise instructions, reviewing output critically, and doing QA like a hawk. If anything it demands more discipline, not less, because you can't rely on catching mistakes by reading your own code you have to actually test everything.

For others who've directed an AI agent instead of hand-coding: has your experience matched that, or did you find a different bottleneck? What ended up being the hardest part for you the instructing, the reviewing, or something else entirely?

How do you actually use AI for long-term planning?

I've noticed something while building products.

AI is amazing at answering questions, but the moment a project stretches beyond a few conversations, things start to break down.

Every new chat means I have to explain:

I've got a function in one of my projects that's been "temporary" for about 4 months now. It works, it's covered by tests, and it's ugly enough that every time an agent touches that file it asks if I want it cleaned up.

Every time I say yes, I regret it a little. Not because the refactor is bad, usually it's genuinely cleaner, but because now I've burned review time on code that wasn't broken, and there's a small chance the agent introduces a subtle behavior change I won't catch until it's in prod.