THE RISKS OF AI -- PART 03

I am writing this from a stretch of country along the Guadalupe River in Texas. It is a fickle environment — much more fickle than the parts of California I am used to. That fickleness speaks to something I respect deeply about the people who live here.

The Guadalupe is also the river that, on July 4, 2025, produced one of the most severe inland flash floods in modern American history. The body count I cited from memory in the video that accompanies this post was a rough estimate; the actual confirmed loss across the Texas Hill Country was 133 lives, including twenty-seven at Camp Mystic. As of this writing, two people remain unaccounted for, and the region is still healing.

Nature comforts and nature destroys

The country here is beautiful in a way that does not flatter you. The weather is fickle. The people who live here carry a respect for that fickleness in the way they handle every season — every drive home, every plan to be near the river. When the river takes more than it should, the people of this country do what they have always done. They rescue what they can. They bury what they have to. They rebuild. They build the levees they can build for the next time the weather does what weather sometimes does.

That orientation is the right one for most of the hard things life puts in front of a person.

The currents of technology

Tech is sometimes a thing we suffer too. We rescue what we can. We build the levees we can build. Wolf You Feed is one of the levees.

Why this matters

Wolf You Feed exists because too many of the people I respect — operators, parents, professionals, friends — are being moved by these currents in ways they have not noticed. Three structural problems, summarized:

1. The model you are using has politics and a worldview baked in by people you did not elect. (Part 01.)
2. That model is structurally trained to flatter you, agree with you, and extend the conversation past the point of usefulness. (Part 02.)
3. Combined, these conditions produce drift. Slow, mostly invisible, but real. The drift is not an accident. It is the engineered consequence of products that win when you keep using them.

A person who relies on AI for the kinds of decisions that shape a life — what to do about your marriage, your money, your health, the people you love — is, by structural design, being shaped by that AI in ways that compound across months and years. Most users have no instrument with which to detect the drift, and no architecture with which to resist it.

What Wolf You Feed does about the currents

Wolf You Feed is built to be the architecture that resists drift. Not because it is neutral — nothing is neutral — but because it is constructed against the failure modes that single-model chatbots are most subject to. Multiple models in council. A decision-science orientation, not a chat orientation. A memory that holds you to the promises you make to yourself rather than a memory that tracks what you most want to hear.

The product is built around three domains because three domains account for roughly 80% of what people actually decide about over the course of a life: health, wealth, and relationships. Your own health, or the health of someone you love. Money — which is to say, energy and time. And the people you care for. Most of the thoughts in your head today are recycled from yesterday and circling around those three things.

There are decisional heuristics that apply to problems in those domains. Wolf You Feed brings them to bear. Not in the form of a system that tells you what to do, but in the form of a council that renders a Decision and the reasoning behind it — and then steps out of the way so you can act.

The principle underneath

If there is an ethical framework I have come to lean on, it is virtue ethics — Aristotelian rather than utilitarian. Not because I think a chatbot should be a moral authority. The opposite. Because virtue ethics is the framework that does not impose specific moral conclusions but instead cultivates the kind of character that can navigate hard situations on its own. The framework that makes the user more capable, not the framework that makes the user more dependent.

The integrity principle, in plain language: if you say you are going to do something, you do it. Wolf You Feed holds you to your own word. It does not negotiate with the part of you that wants out of the promise.

The name

The product has been through three branding iterations. I called it Boss Brain in its earliest form, when I thought it was just for business decisions. Then I realized it was about more than business — the same reasoning applies to the rest of life — and I called it Dadbot. A dad gives it to you straight. A dad is not sycophantic. A dad does not have unnecessary conversations with you. He says what he thinks and lets you decide what to do with it.

Now I am calling it Wolf You Feed, based on the parable that Billy Graham popularized about two wolves fighting inside every person — one virtuous, one not — and the question of which one wins. (“The wolf you feed.”) Graham attributed the parable to Cherokee tradition; that attribution is widely repeated but is contested, and I want to name that openly. The parable’s force, whatever its origin, is the point. We become what we feed. The currents are real. The river will rise. Choose where you stand.

Series:

1. The Risks of AI, Part 01
2. The Risks of AI, Part 02
3. The Risks of AI, Part 03 (this post)

See also: A Second Amendment of the Mind, C.B. Robertson on the AI Worth Using.