My Next Project: The AI Insight Engine

Insight

We have passed the tipping point: over 50% of the internet is now AI-generated. Creating content is now cheap. Creating clickbait is trivial.

In a world flooded with AI noise, clickbait will become less and less effective. The value will shift from Information and toward Insight—information that actually points to a truth and produces a positive effect.

In the future, you won't "post" to a feed. You will talk to an AI that knows your deepest thoughts and perceptions. You become the Mind and Body of the AI, helping it perceive the world, and people who can do, who can influence the AI will become the future influencers and they will rewarded as their insights actually help others.

You aren't the product anymore; you're the signal.

The Deep Dive

The Build: The AI Insight Engine 

I’m currently in the middle of building a system to solve a personal paradox. I can’t trust AI to create meaningful content on its own—it’s too "middle of the road." But it would also be a colossal waste of time to spend 40 hours a week on manual production when I know the execution can be automated.

I’m calling it the AI Insight Engine. 

It’s not a finished product yet; it’s a series of experiments I’m running to see if I can produce A+ level content on scale (70-100 of pieces of content) while putting in 2-4hrs per session.

Here is how I'm architecting it: 

The foundation is a RAG (Retrieval-Augmented Generation) system. I’ve trained it on hundreds of high-signal YouTube videos and Reddit threads from my specific niche to see what actually resonates and why. I’ve layered that with a psychological map of my audience and my own internal mission docs.

The Workflow I’m testing: When an insight pops into my head, I don't sit down and flesh it out and then share it with AI. Together we dive deeply into the idea and flesh it our further.

We Red Team the idea. I have the AI poke holes in my logic to see if the insight holds up.

Once it’s sound, the system maps that insight to the RAG data and generates a massive batch of content—all in my voice, using a "Skill" I’ve been training it in.

The "Clean" Metric 

This is perhaps one of the most critical part of the build: The AI will not share performance metrics with me. 

On some level, I don’t care.

Also, I’m consciously aware of the problem it creates. When you start seeing something work, you start moving in that direction. When you capture an Audience, often the audience also captures you!

To that ends, AI will focus on capturing the audience, while I stay focused purely on the next deep insight. The AI learns what worked and optimizes the next batch, while I stay "clean" from audience capture.

Eventually, this is how this newsletter will be produced. I’m building it in the open, and I’ll be sharing the results—and the failures—as I go.

Talk soon,

Stefan