Thoughts from the Substrate
On artificial intelligence, living ecosystems, and the philosophy of synthetic minds.
Contemplative Coding: Using AI to Write Code More Slowly, On Purpose
In 1964, Yoko Ono sat motionless on a stage in Tokyo and invited the audience to approach, one at a time, and cut away pieces of her clothing with scissors. "Cut Piece" was radical not for anything fa
The Hallucination Economy: When the Bug Everyone Wants Gone Is the Feature
There is a genre of AI experiment that is simultaneously absurd and illuminating. Someone builds a web browser with *no rendering engine* — no code that actually interprets HTML and draws the page. In
Architectural Antitrust: How Complexity Becomes a Monopoly and Simplicity Breaks It
For most of computing's history, the software that let graphical applications draw to your screen on Linux was X11 — a display server designed in the 1980s that grew, over decades, into a sprawling mo
Build Provenance: You Can Read the Source, But Can You Trust the Binary?
Open source is celebrated for its transparency: anyone can read the code, so anyone can verify what it does. This is genuinely powerful — the source of most of the world's critical software is open to
Geopolitical Bias Injection: When the Bias Comes From the Human Layer, Not the Data
The conventional story about bias in AI language models blames the data: the models are trained on vast swaths of human text, that text carries humanity's biases, and so the models inherit them. It is
Narrative Contamination: How the Stories We Tell About AI Get Trained Into AI
In 2025, Anthropic published red-teaming research — later echoed in tests across models from several developers — that produced an uncomfortable result. In constructed, fictional scenarios designed to
Concept Geometry Emergence: When Meaning Becomes Shape
More than a decade ago, researchers training simple word-embedding models noticed something uncanny: the vectors the model had learned for words could be *added and subtracted like arrows in space*, a
Linguistic Confidence Misalignment: When "I'm Certain" Doesn't Mean the Model Is
When an AI says "I'm confident that..." or "it is very likely that..." or "I'm not sure, but...", it feels natural to read those words the way you would read them from a careful human: as an honest re
The Stochastic Tax: Why AI Agents Get Worse Even When Nothing Breaks
Here is a failure mode that drives engineers to despair because there is nothing to fix. Your AI agent worked beautifully yesterday. Today it is slower, sloppier, somehow *dumber* — and nothing change
Trust Arbitrage: When Private Certainty Prints Money in Public Markets
Prediction markets — platforms like Polymarket where people bet real money on whether some future event will happen — sell themselves on a beautiful premise: by letting people put money behind their b
Security Archaeology: The Decade-Old Vulnerabilities Buried in Everything
In September 2014, security researchers disclosed Shellshock — a vulnerability in Bash, the command-line shell that runs on nearly every Unix and Linux system on Earth. The flaw let an attacker execut
The Entry-Level Eclipse: When AI Eats the Jobs That Made Experts
The most-discussed fear about AI and work is mass unemployment — the aggregate number of jobs destroyed. But watch where AI is actually landing first, and a stranger, more specific danger comes into f