00.0Writing

From the notebook

Working notes from across the framework — published as they settle.

A confidence score is not an answer
classificationuncertainty

A confidence score is not an answer

Your model says 90%. It does not say what kind of 90% it is.

Machine learning / decision theory12 min
What companionship requires
encounteropacity

What companionship requires

A companion that can be fully predicted is no longer a companion. A companion that cannot be governed is no longer safe.

AI companionship / philosophy of emotions10 min
Why LLMs reason without rules
stabilizationpolicy

Why LLMs reason without rules

No beliefs. No stored rules. No explicit logic. Yet the output is structured.

Formal logic / AI10 min
How models borrow and what they change
constraintfit

How models borrow and what they change

The template arrives. It does not work yet. Something has to be built.

Philosophy of science / evolution10 min
Where values enter science
upstream-choiceresilience

Where values enter science

Before the data arrives, the choices have already been made.

Philosophy of science10 min
What you cannot see is not what is hidden
opacitylegibility

What you cannot see is not what is hidden

A camera does not fail to photograph infrared. It was never set up to.

Philosophy / AI ethics10 min
Why you cannot build the first word alone
distinctionshared-meaning

Why you cannot build the first word alone

An octopus can solve a maze it has never seen. Its innovations die with it.

Cognitive science / philosophy of mind12 min
49.9 million people in the wrong box
classificationremainder

49.9 million people in the wrong box

When the form doesn't fit, it's not the person who is wrong. It's the form.

Social science / measurement10 min