User Experience Is Computation (madebynathan.com)

The piece reframes every user interface as parallel computation across heterogeneous substrates — most obviously a CPU and a human brain, and in modern settings also GPUs/AI models. The UI is not just presentation but a message-passing protocol: the CPU renders, waits, validates and reacts while the brain perceives, interprets, decides and emits actions. Examples range from a simple “save? (y/n)” dialog to a 60Hz game loop (CPU handles physics/rendering; the brain runs slower, higher‑level prediction and strategy) and an AI chat interface where GPU (model), CPU (I/O) and brain all compute in parallel. Conceptually join points (pressing Enter, submitting a form) are synchronization barriers where substrates exchange state. For AI/ML practitioners this matters because it treats human-users as first‑class computational nodes whose latency, error modes and architectures must be modeled, benchmarked and coordinated. UX becomes “programming for the brain”: layout, wording and motion are opcodes; design systems function like reduced instruction sets to minimize cognitive translation. Practical implications include designing minimal primitives, exposing substrate state (spinners/progress), minimizing join‑point latency, and treating end‑to‑end tests as simple brain VMs. Universal Causal Language (UCL)-style protocols could formalize coordination between humans, silicon and models, enabling explicit multi‑substrate interfaces, better human-in-the-loop systems, and more robust accessibility and evaluation metrics.