the AI strategy that backfired spectacularly
(confession inside)
I'm ashamed to admit this… but here we go :) When I first launched my YouTube channel, I was absolutely CRUSHING it.
My growth was doubling every 3 weeks. If the trend continued, I'd hit 100K subscribers by the end of the year.
Not too bad for a random guy with a mustache teaching people how to vibe code with AI!
Then Life Happened
I had to move out of my house for a remodel (first world problems, I get it).
My entire life was in boxes. My recording setup was a disaster. My routine was obliterated.
And right on cue, my channel performance tanked. Views dropped. Subscription growth fell off a cliff. I was spiraling. So I did what any "smart" person building in AI would do:
I turned to Claude, ChatGPT, Gemini - and asked them to figure out what the hell was going wrong.
"Analyze my data. Tell me why this is happening. Here's all of my competitors. Here's a deep research report of AI trends. Give me a strategy to fix it."
It gave me dozens of new ideas:
- "Better" content pillars
- New script templates
- Algorithm hacks
- Thumbnail strategies
- Giant lists of "what not to do if you ever want another viewer ever again"
And I implemented them fully, no questions asked. You know what happened? My performance got even WORSE. Here's what I learned the hard way:
AI is incredible at generating ideas and hypotheses. But it's absolute garbage at actually THINKING. All it's really doing is predicting what response back to you might sound plausible.
I go through why that's the case in my most recent video:
The Trap Most Vibe Coders Fall Into
It can't tell you which ideas are brilliant and which ones will drive you off a cliff.
When you outsource your thinking to AI, you're basically driving with your eyes closed and hoping the GPS is right.
(Spoiler: it doesn't even know it's driving on a road) This is the exact trap most vibe coders fall into. You ask AI to generate code, features, entire products... but you're not critically evaluating whether those solutions are actually RIGHT.
You end up with Frankenstein apps that screenshot well but collapse under real use.
The difference between builders who ship and builders who spin their wheels isn't access to better tools.
It's having a system for filtering AI's output in order to build WELL.
The PRO Framework
That's what my PRO Framework does:
- Plan = you think first (before touching Claude Code/Cursor/Codex/ETC)
- Refine = you work WITH AI (not for it)
- Orchestrate = you execute with clarity (knowing what matters)
If you're tired of having great ideas that fall apart come implementation time…
And you want to start shipping products that people can actually USE...
Then grab the free PRO Framework primer inside my Skool group.
(If you're on this email list, you're probably in the group already)...
It's a systematic approach that separates real builders from prompt tourists. Grab it free inside Skool.
Let's build something that matters. Sean