I can deterministically access any point in a 2^47 discrete space with constant cost, without traversal, memory, or feedback. The system uses a structured index mapping to define the access order over the space (internally using a PRNG-based permutation). Implementation is contained in a compact module (~500KB .xlsx) that includes the full generation logic (PRNG, ranking, and deterministic index mapping). It does not use external data and is not a database. The generator operates independently of evaluation, while evaluation is performed by a constraint function acting strictly as a filter (accept/reject). There is no feedback loop, adaptation, or learning between generation and evaluation. Status Current prototype is a 500KB .xlsx file managing a space of 2^47. The architecture allows scaling to 2^61 with negligible increase in logic and constant access cost. Not yet experimentally validated at that scale. Question Are there known models that involve: - structured deterministic sampling in discrete spaces - deterministic transformations of indices for exploration control - strict separation between generation and filtering - and absence of feedback from evaluation back to the generator, as an alternative to traversal-based exploration