Minimize total cost.
Use the cheapest adequate path, counting money, latency, privacy loss and switching friction.
About / Ximing Zhou
I am interested in systems whose technical architecture quietly determines who has power.
Artificial intelligence is the clearest example. A model can appear universal while its weights, memory, compute, identity layer and rules remain controlled by a small number of institutions. Cogvo is my attempt to decompose that stack, test alternatives and build tools that give users meaningful agency.
My way of working comes from physics: identify the state variables, constraints and conservation laws before arguing about outcomes. Computer science makes those ideas executable. Political economy asks who owns the infrastructure, who captures the value and how the system changes when incentives change.
I also train as a swimmer. The practice is different, but the epistemology is similar: measurement, feedback, error correction and disciplined repetition.
Operating principles
Use the cheapest adequate path, counting money, latency, privacy loss and switching friction.
No company, model or protocol should become a permanent dependency.
Interfaces, models, retrieval, compute, tools, memory and payment remain modular.
Continuously match each task to the best open system or lowest-cost qualified chain.