# Cognitive Landscapes ### AI Architecture as a Design Challenge for the Harvard Graduate School of Design --- Artificial intelligence is not merely a model. It is an **ecology in which representations, constraints, and actions circulate** — through channels, reservoirs, gates, and transformations. This is not a metaphor borrowed from design. It is the same problem designers have always faced. GSD works at every scale at which flow is governed: the building that organizes human movement, the city block that structures social interaction, the infrastructure corridor that shapes regional settlement, the watershed that determines where life is possible. Landscape urbanism, urban design, architecture, planning — these disciplines share a common object: the design of environments that channel, buffer, distribute, and transform flows of people, water, energy, capital, information, and care. Cognitive landscapes are next. --- For four centuries in the American Southwest, acequias have governed the flow of water through high desert communities. Private landowners. Shared infrastructure. The mayordomo coordinates the flow. The parciantes — the members — govern the commons together. No corporation owns the water. No state agency controls it. Private property flourishes because the commons is collectively designed and maintained. Elinor Ostrom won the Nobel Prize documenting what acequias have demonstrated for centuries: commons governed by those who depend on them outperform both privatization and state control. The architecture determines the outcome. Agentic AI is being born now. Agents that act, coordinate, and run continuously will require access to the most intimate data of daily life: cameras, microphones, location, biometrics, the patterns of how people live in their homes, streets, and cities. The question is not whether this infrastructure will be built. It will be. The question is **who designs it, and for whom**. If agents run in the cloud, data flows up. Intelligence lives on their land. Citizens become tenants — surveilled, extracted from, dependent. If agents run at the edge — on devices people own, federated with neighbors into governed commons — data stays where it is. Intelligence comes to people. They remain owners. --- **What flows** in a cognitive landscape: signals, attention, memory, permissions, goals, messages between agents, actions on the world. **What structures those flows**: models, context, policies, infrastructure, latency, bandwidth, human oversight, social and institutional rules. Donella Meadows identified paradigm — the shared assumption from which a system arises — as the second deepest lever for changing any system. Better privacy policies, antitrust enforcement, and regulation all operate at the shallow end, adjusting parameters while the paradigm holds. **The paradigm-level question is: where does computation live?** On their land, or yours? The six digital common pool resources — compute, storage, bandwidth, power, availability, and accessibility — are finite, they flow, and they must be allocated. No single household has all six in abundance. Federated together, governed collectively, they provide what no individual can and what no platform should own. --- GSD is uniquely positioned to take this on. Every design discipline at GSD already works with the governance of flow at scale. Landscape urbanism designs the infrastructural ecologies through which cities metabolize resources and waste. Urban design structures the channels through which public life moves. Architecture organizes the thresholds, gradients, and enclosures through which people inhabit space. Planning governs the rules by which all of this is built and maintained. Cognitive landscapes extend this tradition into the flows that now define daily life as completely as water, movement, and energy once did. The design challenge is to design the **channels, gates, and governance structures** through which cognitive flow moves through communities — such that intelligence comes to people rather than people being delivered to intelligence. This is not interface design. It is not model optimization. It is not policy. It is the design of inhabited infrastructure at the scale at which GSD already operates, for a medium that is reshaping every scale at once. The design question: what does it mean to build a city, a neighborhood, a territory, a commons, in which AI works for the people who live there — governed by them, serving their land, learning from their patterns without surrendering those patterns to extraction? That is a GSD question. It always has been. --- *The water is flowing. The question is whether we will design the ditches ourselves — or let the technofeudal lords build the dams.*