The Self-Creating Universe framework proposes that the fine-tuning problem is resolved not by invoking a creator, multiverse, or external simulation, but through selection by mathematical necessity. The universe is treated as a zero-flux boundary eigenproblem: a standing-wave solution whose physical constants are determined by the simultaneous satisfaction of past inflationary constraints and a future intelligence-viability constraint, phase-locked by a conserved informational invariant. Only universes capable of generating intelligence that can retrocausally stabilize their own initial conditions achieve self-consistent causal closure and therefore exist. The causal loop is structural rather than temporal — it may have already closed elsewhere in the universe's history — and is grounded physically in Aharonov's Two-State Vector Formalism, experimentally confirmed through weak measurement protocols. The "boundary determines bulk" principle is supported by the AdS/CFT correspondence. The learnability of physical law is explained mechanistically through Zurek's Quantum Darwinism. The Boltzmann Brain objection is resolved by a double exclusion criterion: Boltzmann Brains satisfy neither the future intelligence-viability constraint nor the Quantum Darwinism reliability condition. The causal loop's non-temporal character is formalized through Causal Set theory. The intelligence threshold is operationalized via Tononi's Integrated Information measure Φ. The framework is distinguished from Wheeler's Participatory Universe, Tipler's Omega Point, and Tegmark's Mathematical Universe Hypothesis, and operates as a selection filter within the latter. Six explicit falsification criteria are provided. Testable predictions include spectral rigidity in ratios of fundamental constants, scaling of quantum retrocausal effects to macroscopic systems, and bounded algorithmic complexity of cosmological parameters.