Information comes first
In its most primitive sense: distinguishable state. Difference. Change. The possibility that one state could have been another.
A relationship can matter as much as the things it connects
In many systems, relationships don't just describe objects — they constrain them, enable them, transform them, give them meaning.
Context changes state
A description can be accurate in one frame and misleading in another. Time, observer, scale, boundary, purpose, environment. Context is not decoration.
Emergence isn't just a side effect
If a recurring pattern gets stable enough for the rest of the system to treat it as a unit, that pattern has become operationally real. A representation should be able to account for it.
The observer is part of the description
Boundaries are drawn from somewhere. Variables are chosen by someone. Measurements go through instruments. Every model contains perspective. Pretending otherwise doesn't remove it.
The map deserves scrutiny
AI spends enormous effort evaluating models. The representations we hand those models deserve the same attention. A powerful reasoner over the wrong map still produces the wrong route.
Complexity must earn its place
We like complicated ideas. We don't worship complicated implementations. If a linear model solves it, congratulations to the linear model.
Claims should carry labels
Belief isn't evidence. A hypothesis isn't a result. A result isn't replication. Replication isn't universal truth. The distinctions matter.
Failure is information
A failed experiment can improve the map. Hiding it can't.
Reality gets the final vote
Always.
We suspect reality is better understood as persistent patterns of changing state than as a collection of fundamentally isolated things. We suspect information, relationship and recurrence are deeper explanatory tools than our software architectures usually admit. We suspect these ideas can improve how intelligent systems represent and reason about complex environments.
Those are significant claims. Some are philosophical. Some may become scientific. Some may turn out to be wrong.
Our job isn't to protect them. Our job is to make them precise enough that reality can answer back.