The form asks for your address before your name. You’ve already typed three characters into the first field before your eyes catch the label. Delete, tab back, reorient. Five seconds. You do it forty times a day, and by 2 PM the tank is empty. The resource that lets you sequence, prioritize, and switch between contexts has been drained by friction the designer never considered.

Most systems are designed under ideal conditions. The designer has full attention, low stress, a clear head. The system reflects those conditions back, assumes the person using it will too. When they don’t, the system breaks. Or worse, the person breaks against it.

I’ve spent twenty-five years building under the opposite condition. Executive function friction, processing limitations, days where the cognitive budget is half what the environment expects. That constraint forced a specific kind of architecture: systems that work when willpower isn’t available. Systems where the decisions have already been made. Systems that carry more weight so the person carries less.

The household version is a batch cooking system I run every Sunday. Four family members, four different dietary restrictions, and some mornings I’m operating at maybe 40%. Without structure, breakfast is a four-variable problem I have to solve before coffee. With the system, the bases were built on Sunday. The assembly is mechanical. Open, portion, heat, plate. The low-capacity morning was the design requirement, and the system that meets it works better for everyone on every morning, including the good ones.

The professional version is the same operation at a different scale. At Encore, a twelve-year enterprise engagement, I watched engineers burn executive function on questions the system should have absorbed. Is this the right template? Does this match the brand? Each question is a five-second tax on someone who should be building, not guessing. I redesigned the system to answer those questions before they got asked. Defaults replaced decisions. Structure replaced judgment calls. The cognitive load dropped, and the error rate dropped with it.

The principle generalizes. A CLAUDE.md file (persistent project context for an AI tool) is an executive function accommodation. Instead of re-explaining constraints every session, you write them once. The tool reads them. Your cognitive resources go toward the actual work. The Savepoint Protocol does the same thing for thinking: it externalizes a decision at the moment it crystallizes, so you don’t hold it in memory across sessions. Working memory has limits. Mark the progress before coherence degrades.

Every one of these systems came from the same place: a processing reality that couldn’t be wished away. The morning system exists because some days the tank is low before breakfast. The voice protocol exists because evaluating “does this sound like me” on every sentence compounds across a hundred pages. The decomposed evaluation system exists because asking a model to assess four dimensions at once produces the same degraded output as giving a student three instructions disguised as one. One objective at a time. One clear output before the next step.

The accommodation methodology (decomposition, scaffolding, single-objective tasks, externalized state) didn’t emerge despite the constraint. It emerged from it. The constraint was the design requirement. Every system I build starts with the same question: where is this asking more of the person than it needs to? The answer is always structural. Remove the load. Build the defaults. Let the person do the work that actually requires their attention.

Systems designed for the worst day work better on every day. That’s the principle. The constraint isn’t a deficit to overcome. It’s a processing reality that, when you design for it honestly, produces architecture that holds for everyone.