Avoiding Star Worship and My Bullshit Detector
Back to Blog
AI Philosophy

Avoiding Star Worship and My Bullshit Detector

Pete Gypps
Pete Gypps
Published: 5th July 2026
5 min read

It all starts with the foundation.

Not the AI, not the data — a rock-solid foundation and a thesis we've tested to death and actually trust. That comes first. Everything else hangs off it.

Aligned intake

Because once the foundation's solid, you know what you're looking for. You don't vacuum up the whole internet and hope. You go looking for the stuff that aligns with the thesis — and that's the biggest filter of the lot. Most of what's out there isn't even wrong. It just doesn't fit how we build. Not for us.

Then you sweep what's left. Two passes.

First: star worship

Someone's got a hundred thousand followers, so they must be right? They're not — they're just loud. A take gets repeated until it looks like consensus, but a thousand people quoting one bloke is still one bloke. Reach isn't correctness.

Second: the bullshit detector

Same question, sharper: is this actually right, or does it just sound right? Those two came apart a long time ago, and AI blew the gap wide open — it hands you what it knows and what it made up in the exact same confident voice.

And that's the whole point. Foundation first, aligned intake, then the sweep. That chain is the only reason any of it can grow on its own. Autonomous growth isn't the AI being clever — it's a foundation solid enough that new stuff either fits it or gets thrown out, with nobody standing over it. Take the foundation away and "grows autonomously" just means "drifts autonomously". That structure-first foundation is my AI First Principles framework.

The sweep has to doubt itself

Once, the detector flagged a claim from a video — that a researcher had joined a particular lab — as wrong. Confident. Reasoned. But a flag is itself a claim, so it had to check a live source first. Good job: he'd joined the day before the video went up. My training data had a cutoff; reality doesn't. The detector's own confident "correction" was the actual bullshit — caught before it grew into anything.

Better models won't save you here — smoother bullshit, faster. You can't outsource the sweep to the thing you're sweeping.

So the foundation holds, the intake aligns, the sweep stays honest — and the thing grows without me standing over it. That's the whole game: build the foundation solid enough that growing on its own is safe, not scary.

Pete Gypps

Written by

Pete Gypps

Founder & Solutions Architect

About This Article

Most of what is out there is not even wrong — it just does not fit how you build. Foundation first, aligned intake, then a two-pass sweep: star worship, then a bullshit detector that has to doubt itself too.

Let's Connect

Have questions about this article or need help with your IT strategy?

Book a Consultation
P
Pete Bot
Business Solutions Assistant
P

Let's Get Started!

Enter your details to begin chatting with Pete Bot

💬 Got questions? Let's chat!
P
Pete Bot
Hi! 👋 Ready to boost your business online? I'm here to help with web design, SEO, and AI solutions!