Modern distributed systems address computation by location: hostnames, IP addresses,URL paths. We argue that this is a structural error, not a tooling gap, that forces protocol fragmen-tation, prevents cloud-edge unification, and makes systems inherently brittle. The emergence of AIworkloads that must execute across data center GPUs and microcontrollers with 400 KB of RAMmakes this error untenable.We propose intent-oriented computing, a paradigm in which the unit of distributed invocationis a semantically-addressed, version-resolved, quality-of-service-annotated intent that carries noreference to any physical endpoint. We define the intent abstraction, enumerate seven categories ofinfrastructure it eliminates as separate concerns, and present evidence of feasibility from a completeimplementation spanning 200,000+ lines of Rust and 20,000+ lines of C, running across cloudservers and ESP32 microcontrollers.This paper introduces a paradigm. A production-grade implementation exists and validates theclaims herein, but the contribution is the abstraction, not the artifact. It is intended as an invitationfor the systems community to explore post-address computation.
Vinícius Queiroz de Albuquerque (Thu,) studied this question.