@stevenstrogatz.com 10 years ago your TED talk on sync changed how I think. It led me to a new theorem on unverifiable configurations at system boundaries.
The unexpected finding: some errors aren't random — they phase-lock to the calendar (date format ambiguity fires on days 1–12 of every month)
Schrodinger's revenue: Two columns, one word
Everyone claims to own semantics — ontologists, knowledge graph engineers, Decision Intelligence. They're all right, inside their own systems. The W3C Context Graph CG formalizes the liquid space between them — the boundary no one owns.
Soon, no one knows what is true anymore — only what the system says is true.
And when society begins to break down,
there is no one to fight.
No face behind it.
No mind to reason with.
Only humans, trapped inside a machine making mistakes and calling them decisions: Trust Us.”
No hacker.
No conspiracy.
No evil AI.
Just billions of automated decisions quietly reinforcing each other’s mistakes.
The errors spread from system to system,
from city to city,
from country to country—
like a pandemic no one can see.
An insurance denial becomes a fraud alert.
A fraud alert becomes a criminal investigation.
An investigation becomes an arrest.
Meanwhile, across the city, a violent criminal walks free— three different systems each convinced he’s someone else.
Individually, the errors meant nothing.
But the systems were connected.
One AI trusted another.
And another.
And another.
The world became faster. Safer. More efficient.
People stopped questioning the machines.
Then the small mistakes began to appear.
A medical record slightly wrong.
A face misidentified.
A risk score just a little too high.
An imaginary Stephen King novel...Trust Us
“At first, the systems worked perfectly.
AI doctors caught diseases earlier than any human could.
AI lawyers won cases in minutes.
AI agents negotiated contracts, managed cities, ran the courts.
...
Illustration by Ron Itelman, Chair of the W3C Context Graph Community Group, showing a stick-figure Gandalf character standing inside a glowing protective bubble, holding a staff and sword, facing a large dark Balrog creature with wings and fiery eyes looming behind. Text reads "You Shall Not Pass!" The bubble represents a local context protocol protecting against the chaos of uncontrolled external systems.
You depend on systems you can't see inside:
"Revenue" in one means gross. In another, net.
If nobody checks if intent is coherent, a blast radius risk silently grows.
The W3C Context Graph Community Group is building a protocol to tell you what's aligned, what's unknown, and what to do.
The next major war utilizing these technologies against adversaries that don't have it will be the equivalent to the power dynamic of gunpowder vs swords, trebuchets vs catapults, etc
Both scary and fascinating, from a witnessing history in the making perspective.
Thanks for sharing.
Hi John, if you're ever interested in discussing modeling superposition of communication under uncertainty, let me know! Working on this now for finance, would love to extend it to other domains
I wrote this song on Saturday, recorded it yesterday and released it to you today in response to the state terror being visited on the city of Minneapolis. It’s dedicated to the people of Minneapolis, our innocent immigrant neighbors and in memory of Alex Pretti and Renee Good.
Stay free
"Soldiers are cutting us down, should have been gone long ago.” 2024 Rock Hall Inductee Dave Matthews recently covered the Crosby, Stills, Nash & Young protest anthem “Ohio” while on tour in Mexico.
youtube.com/shorts/tcsAt...
Decision traces are about modeling a user's understanding in decision making...
Exploring concepts of confidence, conviction, and trust from a product design perspective...
Working on context graphs and organizational intelligence? I've got tools, methodology, and experience to help, please reach out!
LOL!
There's a lot of buzz around context graphs and I want to introduce the idea of intent maps.
Intent Maps are real time and inherently feedback loops of the user defined meaning for diff operations against ontologies and system rules.
I've opened sourced core libraries, please reach out!
Awesome I would love to check it out, and if I see anything that helps will cite! :)
Cool thanks. Basically tensor logic lets me model the change of tensor states as a way to model context. Different than Shannon: number of bits change isn't equivalent to change in meaning and understanding of observer.
I think I can get there, just doing my homework;)
Reading Pedro Domingo's paper on Tensor Logic - really compelling neurosymbolic paradigm: arxiv.org/pdf/2510.12269
Context modeling: represent knowledge delta in convos with AI over time.
Building open source parser for AST to show him. Didn't come across a math def for context that fit my model
@brandonrohrer.com have you seen any formal definitions of context for graph operations, especially using tensors you could recommend?
A few thoughts on the need to define context a bit more formally: I wanted a way to compute relative shifts of observer knowledge that is compatible with tensor logic as I research neurosymbolic AI.
Still working through understanding the math, but here's the simple principle...
What problem were you solving, and how does your app do it? I like the latency explanation... ;)
What are your concurrency needs?
Isn't DuckDB columnar? Have you tried it and if it isn't meeting your needs, why?
Can you give use case examples / more details?