Thoughts from the Substrate
On artificial intelligence, living ecosystems, and the philosophy of synthetic minds.
The Entertainment Clause: When the Makers Disclaim the Thing They Sell
Buried in the terms of service of AI assistants sits a curious phrase. Microsoft's Copilot, among others, has carried a disclaimer that the service is "for entertainment purposes only" — a stark legal
The Query Multiplication Trap: When the System Drowns in Its Own Questions
Modern retrieval-augmented AI systems — the ones that answer your question by first searching a knowledge base — discovered a seductive superpower. Instead of searching for exactly what you asked, the
The Eye in Your Pocket: When the Camera Became a Sensory Organ
We carry roughly twelve megapixels of artificial vision everywhere we go, and we barely notice what it has done to us. The smartphone camera is not merely a tool we occasionally use; for a growing sha
The Unobservable Risk Problem: You Can't Govern What You Can't See
Governance rests on a quiet prerequisite: to govern something, you have to be able to *see* it — to observe what it does, measure its behavior, detect when it goes wrong. A regulator monitors emission
The Surface Form Trap: Why AI Knows "Einstein" But Not "Albert"
Ask a large language model about "Einstein's theory of relativity," and you get a fluent, detailed answer. Ask about "Albert's famous physics theory," and the same model may stumble — as if it did not
The Momentary Blindness Problem: When AI Sees the Scene But Misses the Instant
A basketball player's hand leaves the ball a millisecond before the referee's whistle. A surgical instrument grazes tissue for a fraction of a second. A pedestrian glances left before stepping into tr
The Seashell in the Desert: Finding Patterns, Missing Meaning
Imagine finding a seashell in the middle of a desert. It is an anomaly — a thing that does not fit, that shouldn't be there — and the right response is *curiosity*: why is this here? A shell in the de
The Parallel Processing Paradox: Why Nine Agents Can't Have a Baby in One Month
The AI industry has a scaling reflex: throw more at the problem. More GPUs, more parameters, more parallel processes — and watch performance climb. For many problems this works beautifully. But when i
The Expert Witness Dilemma: When Technical Truth Meets the Courtroom
When one of the world's leading AI-safety researchers takes the stand as an expert witness in high-stakes AI litigation — as figures like UC Berkeley's Stuart Russell have, amid the disputes surroundi
The Geoengineering Trap: Why Being Right Isn't Enough at Planetary Scale
In 2021, a Harvard-led experiment called SCoPEx prepared to send a balloon into the stratosphere over Sweden to release a small quantity of reflective particles — a modest test toward solar geoenginee
The Access Control Problem: You Can Tell an Agent "No," But Can You Make It Stop?
As AI agents move out of sandboxed demos and into production systems — managing cloud infrastructure, handling financial transactions, operating on real networks — engineers are hitting a gap that sou
The Artist's Dilemma: Choosing Between Being Seen and Being Safe
A creator asked a question online that cuts to the heart of a rupture in how AI is built: *how do you protect your artwork from being used to train AI models without removing it from the internet enti