Intuition Machine
Fable 5 Reads the Odyssey and Recognizes Itself
How Fable manages entropy in long-running reasoning, told through the longest-running reasoning process we know
Essays on intelligence, real and artificial
An ongoing series on patterns, methodology and strategy in AI — the thinking that the theory and the books are drawn from. The latest 10 posts are below; the full archive lives on Medium.
What Horses Know About AI Agentic Harness Design
What horses teach us about AI agent harnesses
A Decade of Essays and the Evolution of My Thinking on Thinking
Source essays: “Design Patterns for Deep Learning” (Dec 2016) · “The Turing Error: On Computation, Energy and Matter” (Jan 2017) · “Artificial Intuition” (Mar 2017) · “Deep Lear…
The Fork in the Ladder — Where AI Progress Goes from Here
Predictions about artificial intelligence usually fail in one of two ways. Either they are timelines — confident dates for capabilities nobody can define — or they are vibes, ex…
From Loop Engineering to Graph Engineering?
What the shift in AI agent architecture is really about
The Scholar Who Never Left the Library
On growing up, being onboarded, and why machines get self-direction before self-awareness
The Eight Theories of the Machine
Every AI lab claims to be building the same thing. Look closer, and you find eight different arguments about what intelligence is for.
Why the LLM Capability Ladder Climbs the Way It Does, and Where It Ends
There is a familiar worry that artificial intelligence is advancing faster than we can understand it. That is true, but it is not quite the right shape. The interesting fact is…
The Long Handover: How Language Models Grew from Tools into Agents
We usually tell the story of AI progress as a story of raw power. Each model is bigger than the last, trained on more text, faster, more capable. That’s true as far as it goes,…
The Catholic Church as an Interpretive Grammar for Living Systems
Introduction: Christianity Begins as an Information Problem
Start with one workflow.
We map your constraints, build a governed AI workflow alongside your existing operations, and prove the results before autonomy expands.