Cognitive architecture treats human attention as infrastructure — it defines the boundary between what systems must absorb and what people must still be able to understand
These are field notes from architecture at scale. Not opinions, not tutorials, but observations distilled from repeated contact with large systems and their constraints.
##
IDX · NOTES
01
These are not beliefs or preferences. They are the conditions under which systems either hold or fail.
§ Complexity migrates
Complexity does not disappear. It migrates. Remove it from architecture and it appears in people. Hide it in interfaces and it accumulates in process. Architecture determines where complexity resides.
§ Cognition is finite
Human attention is limited. Data is not. Systems that ignore this externalize cost into errors, delays, and manual overrides. Cognition is a hard constraint.
§ Automation is not intelligence
Automation is execution without understanding. Intelligence requires explicit constraints and bounded context. Without them automation simply accelerates noise.
§ Control without continuous attention
Control that requires continuous attention is not control. It is an unstable equilibrium. Good systems hold on their own and spend attention on meaning, not maintenance.
§ The system must explain itself
A system that cannot explain its state cannot be considered reliable. Explainability is part of architecture, not external documentation.
02
Complexity does not disappear. It migrates.
Every attempt to "simplify" a system moves complexity somewhere else. The only real choice is where it will live.
§ Complexity migrates
When complexity is removed from architecture, it reappears in people. When it is hidden behind interfaces, it accumulates in process.
When it is pushed out of code, it surfaces in meetings, escalations, and exceptions.
§ Architecture as pinning
Organizations often mistake relocation for elimination. They celebrate cleaner UIs, fewer controls, faster flows — while silently increasing cognitive load downstream.
This is why many systems feel easy to build and hard to operate. Architecture is not the act of removing complexity. It is the act of pinning it down.
§ Localizing complexity
A well-designed system makes complexity explicit, localized, predictable. A poorly designed one lets it diffuse — until no one can point to where decisions actually happen.
§ Weak boundaries
Complexity migrates toward the weakest boundary. Architecture decides where that boundary is.
03
Human cognition is finite. This is not a preference. It is a limit.
Data, events, metrics, and correlations are not bounded in the same way. Any system that treats these two as symmetrical will eventually fail its users.
§ Cognition is finite
Most enterprise platforms fail not because they lack intelligence, but because they assume infinite attention. Dashboards grow. Filters multiply. Context fragments. At some point, the system stops supporting decisions and starts demanding them.
§ Asymmetry of data
This is the moment where automation is usually proposed as a solution. But automation without constraint is only accelerated confusion. Intelligence begins where context is bounded, variables are constrained, and ambiguity is acknowledged, not hidden.
§ Automation and noise
The role of architecture is not to expose more information, but to compress it into a decision surface that a human can actually operate. When cognition is treated as a soft concern, systems externalize cost into errors, delays, and quiet burnout.
§ Architecture as compression
Cognition is not a UX problem. It is a system-level constraint. Ignoring it does not make systems smarter. It only makes failure slower and harder to diagnose.
// notes.cadence
Texts appear rarely. Only when a pattern becomesunavoidable.