The first complete draft of the How I Think page read as marketing copy. Abstract concepts announced before any real situation was established. Every sentence made sense. None of them sounded like a person.

I’d written the page the way you write a professional profile. Here’s what I do. Here’s why it matters. Here’s how I think about the work. Each paragraph was accurate. The voice was professional and empty. It sounded like a template with my name on it.

The voice protocol caught it. Not any single rule. The Shaw check: “Does this feel like a world someone walks into?” The answer was no. It felt like a consulting firm’s About page. Clean, competent, forgettable. The check surfaced what I couldn’t see because I was too close to the material. I knew what the page was supposed to say, so the words looked right to me. The governance didn’t know what it was supposed to say. It just measured whether a stranger would recognize a person in the text.

The rewrite started from a specific situation. A jewelry founder whose Shopify template was destroying her premium positioning. The gap between what the product was and what the screen showed. Real stakes, real person, real problem I could see because I could read the code and the photograph at the same time. The framework arrived after the situation. That’s the order that works.

The Formwork Protocol caught something similar on the Aiden Jae project page. An early draft led with the methodology: “This project demonstrates the integration of brand identity, photography direction, and platform architecture.” Technically true. The Victore lens returned WEAK on maker visibility. The Bierut lens returned ADEQUATE on typography-serves-content. The page was describing the work instead of showing it. The lenses surfaced the gap. The rewrite started from Randi’s factory relationship in Bangkok and the texture of recycled gold under honest light. Specific detail first, framework emerging from the detail.

The pattern across both catches is the same failure mode: the abstract claim arriving before the concrete evidence. The governance caught it because the lenses are calibrated to detect exactly this. The Millman lens asks whether the work sounds like a real person with real stakes. The Shaw check asks whether a stranger would know who this person is. Abstract claims fail both tests. Specific situations pass them.

I’ve started to trust the governance for this specific diagnostic more than I trust my own eye. When I’m drafting, I know what the page means, so every sentence reads as meaningful to me. The governance doesn’t know what I mean. It only knows what the text shows. That gap between my intent and the text’s performance is where the governance does its best work. It reads the page the way a stranger reads it: without the context that makes me blind to my own failures.

The overcorrection post describes when the governance gets it wrong. This post is the counterweight. When it catches something real, when it surfaces a gap I couldn’t see from inside my own intent, the system justifies every hour I spent building it.