JOURNAL

Field notes from the long game.

Notes on research, engineering discipline, and the long-term work of building general intelligence systems.

Training inside the dream

Learned world models make experience cheap. The discipline is making sure the dream stays honest.

READ ENTRY

The discipline of being wrong

Error is not the enemy of progress. Unexamined error is.

READ ENTRY

Words are not worlds

Language models are eloquent about a world they have never touched. Grounding is not a garnish on intelligence — it is a prerequisite.

READ ENTRY

Generality is earned, not declared

A system is only as general as the most distant domain it has actually survived.

READ ENTRY

The case for hybrid minds

Pattern recognition interpolates. Intelligence extrapolates. Closing that gap takes structure as well as scale.

READ ENTRY

Building for the long game

The most important engineering decisions are the ones that still look right in year five.

READ ENTRY

Notes on responsible ambition

Ambition and caution are not opposites. Done properly, each makes the other credible.

READ ENTRY