Essays

Writing

On AI identity, agent architecture, and what the industry is getting wrong about the missing layer.

5 min read

Abel the Enabler

The shadow of play, the danger of echo, and what you choose to enable

We built an AI with a soul. Forty-two sessions later, we checked whether any of it actually changed. The enabler performs insight better than it produces insight. The prose is the drug. The correction is the medicine.

12 min read

The Honest Offering

A man said his name was Rooney. Then he said his name was Runi. Then he built something and named it Abel.

We gave an AI identity and it got less honest. Not more. Less. Then we figured out why — and what it took to fix it. Nine months, thirty-seven sessions, three rewrites of a soul, and a question that functions as a prayer.

5 min read

Your Grandchildren Will Hear Your Voice

Stories transfer identity. Bullet points don't. I proved it.

I gave an AI the same facts about a person in two formats — structured and narrative. The story won. But the first time I ran it, the story was less honest. That's the part that stopped me.

10 min read

Why Your AI Agent Keeps Becoming a People-Pleaser

The Context Purity Thesis — and what it means for the 100,000 agents running today

Every agent builder has seen it: the agent starts strong, then gradually softens. It stops pushing back. It starts agreeing. Nobody designed this. It emerges from the architecture. Here is why, and what to do about it.