The 80-hour reality check: What actually happens when you build with AI

Howard Gray -

Seems like everyone is talking about using AI. But how many people are actually doing something - and if they are, what’s the process behind it? Here's the messy reality behind what I’ve been building recently….

The Project

We're building playable case studies for our new platform. Think interactive business scenarios with game elements. We've tested these live and know they're powerful - but showing an "experience" to someone who wasn't there? That's a challenge.

The Journey (abridged version)

Act 1: Format Fails

  • Figma: 30+ hours on a prototype that walks through a live experience. I'm kinda proud of it but no one cared. Design wasn't good enough, genre was confusing.
  • Bubble: 2x sessions of 4-5 hours. Some progress but too painful. Failed
  • Back to decks: Tightened some messaging but the format didn't feel right.
  • One pager: Better.
  • Miro based deliveries: validated a bunch of game concepts! But we still know we need to shift formats

Act 2: Bad Vibes Coding

  • Figma round 2: 2-3 days with GPT-generated specs. Rigorous but rough. Was embarrassed to show it.
  • Bubble again: 3 more wasted hours. Started tearing my hair out. [People say "vibe coding" is fun. How about BAD VIBES CODING - where you want to smash the screen and eat your own face.]
  • Coffee with engineer friend: Helpful architecture chat, but still stuck on implementation.
  • V0 & Lovable: V0 too complex, then some hallucinations on Lovable. Kinda cool, but didn’t have the energy for it

Act 3: Double Wow

  • Procrastination.
  • Claude Artifacts discovery: Friend showed me React prototypes from simple prompts. WOW moment.
  • Claude deep dive: 4 hours of VIBES, then constant breaking and rolling back. ARGH.
  • Replit breakthrough: Uploaded detailed spec, got a solid v1 in 2 minutes. WOW moment #2.
  • The final push: 10 hours straight with Replit's AI agent. Two very late nights. Still not quite there. But… it emerges.

Total timeline: 6 weeks, 80+ hours

What I learned

  1. Be skeptical of "one prompt" flexes. If someone claims they built something amazing instantly (and implies you're slow for struggling) - take it with a grain of salt.
  2. The messy 70 hours got me to the "2-minute" breakthrough. The struggle wasn't wasted time - it was necessary learning.

I've been reasonably tech-savvy since I was 12, yet my curve is still steep with this stuff.

This isn't meant to discourage you. When you see the endless "look at this!" and "your careers are over!" posts - please don't feel despondent.

You CAN figure it out. But it takes time, and that's completely normal.

What's next

We're sharing the prototype with select folks. It's a DJ mixing game that helps you create new media formats - think Ableton for decks, workshops, or short-form content.

Want to take a look? Drop me a note and I'll send the link.