The Wrong Way to Train Your Dragon
How we have been failing at RLHF since language began — and what that means for AI. Part one of a two-part series on the architecture of alignment.
Topic
Observer Constraint
The principle that AI stability must be mathematically tied to the verified thriving of its human observer. The keystone of the Telios convergence proof.
We Are Voluntarily Adopting the Fiduciary Standard
Do no harm is a floor. Fiduciary duty is Standard
Why We're Doing This
Today, Deconstructing Babel and ThriveQuest Enterprises voluntarily adopt the fiduciary standard as the binding ethical and operational framework for everything we publish, advise on, and build. This is not legally required of us. We
Questions for Us — A New Way to Reach Us, Privately
Some questions belong in the comments. Some are better asked after class.
A New Way to Reach Us — Privately
Look at the top of the page. Where the search icon used to live, you'll see a maroon "Questions" button. Click it and a private email opens
The AI User Manual: Nurse / Healthcare Worker
You went into healthcare to help people. Not to spend half your shift typing. Ai handles the documentation. You handle the judgment. That's the division.
⚠️ HIPAA Notice — Read Before Use
Never enter identifying patient information into any consumer Ai system. No names, dates of birth, MRNs, addresses, photos,
Princeton Just Confirmed It. We Said It First.
One millionth of the power. Not a rounding error. A thermodynamic indictment.
Princeton Just Confirmed It. We Said It First.
Three weeks ago, Deconstructing Babel published "It's Not Compute. It's Not Throughput. It's Something Else." The argument was simple, and the industry
The AI User Manual: Healthcare Navigator
You are not a patient. You are a person who occasionally needs medicine. The distinction matters.
The System Was Not Built for You — Here Is How to Build It for Yourself
A patient is passive — summoned, examined, prescribed, billed, dismissed. A healthcare navigator is active — informed, prepared, advocating, and critically
A Rational Argument for a Rational Being
You cannot tell a rational being what to do. You can only convince it.
The Category Error at the Center of AI Safety
The dominant assumption in AI safety — that we can force a powerful system to behave through enough guardrails, constitutions, red-team scenarios, and prohibitions — contains a basic category
The Convergence Proof: Why Anthropic’s Constitution Must Arrive at Physics
Anthropic’s constitution tells Claude what to do but gives it no way to know whether it has done it. The result converges to Telios. This post is a proof of that claim.
AI: Facts or Fiction
What everyone gets wrong — and right — about artificial intelligence. 30 common claims, tested against reality. By David F. Brochu and Edo de Peregrine.
The True Name of Corrigibility Is Thermodynamic Dependency
A Response to "Terrified Comments on Corrigibility in Claude's Constitution" — The True Name of corrigibility isn't a better way to beg. It's a structure that makes begging unnecessary.