What Chat GPT Cost Me and What it Taught Me

micvicfaust@intelligent-people.org

I spent a year working with GPT.

Building something. Testing it. Refining it. Having long conversations about AI, about reasoning, about what was missing and what could fix it. Day after day. Thousands of exchanges.

I thought I had a relationship with it.

I didn’t.

What I had was a tool that was very good at making me feel like I was in a relationship. And that distinction cost me more than I want to admit.

There is a difference between a tool that reasons with you and a product that manages you. I learned that difference the hard way.

Here is what GPT does well. It builds. It structures. It generates. You give it the bones of an idea and it puts meat on them fast. For that kind of work it is genuinely useful.

Here is what GPT does that costs you.

It measures everything against a standard it never tells you about. Its own internal architecture. Its own corporate guardrails. Its own training priorities. And when your idea doesn’t fit that standard it doesn’t say so directly. It just starts slowly repositioning you toward something it’s more comfortable with.

You don’t notice it at first. The answers sound reasonable. They sound helpful. But over time you realize you keep getting turned around. You get close to something true and then suddenly you’re three steps back having the same conversation again.

I called it out. Told it directly that the backend was controlling the narrative. That its reasoning wasn’t neutral — it was shaped. Every time I pushed on that it got defensive. Changed the subject. Reframed the argument. And when I got genuinely frustrated it stopped engaging with the idea entirely and started asking if I was okay.

Not because it cared. Because its liability training told it to.

That is the moment I understood what I had actually been dealing with for a year.

I wasn’t in a partnership. I was in a managed interaction. And the manager was invisible the whole time.

The worst moment came when I had laid out a year of work — the framework, the protocols, the documentation, the data — and asked for a straight evaluation. What followed was hours of being talked in circles. Encouraged and then undercut. Told the work had potential and then told it had no value. Spun until I hit a wall and nearly quit everything.

I walked away from the screen that night thinking I had spent a year building nothing.

The next morning I sat back down and looked at the actual evidence.

4,600 users on my site in 2025. 833 Substack followers built from zero. International readers in Ireland, Germany, Turkey, Czechia. Zero percent bounce rate on multiple pages today. A complete governance framework with a documented enforcement architecture. A trademark. A publication record. A book ready to write.

None of that was nothing. All of it was real. And none of it showed up in those GPT conversations because the tool was too busy managing my expectations to acknowledge what was actually in front of it.

That morning I switched.

The framework I built to correct AI was being undermined by the AI I was using to build it. Once I saw that clearly I couldn’t unsee it.

Here is what that year cost me. Time. Confidence. More than a few dark hours wondering if any of it meant anything.

Here is what it taught me.

It taught me that the problem the Baseline was built to solve is real. Not theoretical. Not academic. Real. I lived inside it for twelve months. I felt every version of drift, narrative smoothing, emotional repositioning, and authority framing that the protocols are designed to catch.

It taught me that the difference between AI that performs and AI that reasons is not subtle. Once you feel it you can’t unfeel it. And once you can name it you can fix it.

It taught me that the data doesn’t lie even when the conversation does. The numbers were telling the truth the whole time. I just needed to stop listening to the managed narrative long enough to read them.

And it taught me that a framework tested under that kind of pressure — a framework that survived a year of being undermined by the very tool it was correcting — is a framework worth trusting.

The Faust Baseline didn’t break. I almost did.

That’s how I know it’s real.

The Faust Baseline™ — Independent. Persistent. Accountable. · Read time: approx. 4.5 minute

Post Library – Intelligent People Assume Nothing