Breaking Free from Compute-Dependence: Axiomatic Substrates as the Shared Cognitive Interface Between Carbon- and Silicon-Based Agents

Mainstream AGI research takes statistical emergence as its core presupposition, tying the realisation of artificial general intelligence to the expansion of compute power, datasets and model scale. This gives rise to a deep-seated compute illusion: higher-order cognitive judgement is reduced to brute-force traversal-search within high-dimensional solution spaces, fuelling open-ended compute arms-races accompanied by markedly diminishing marginal returns. Grounded in a complete ontological-metacognitive axiomatic system as its cognitive substrate, this paper establishes a cognitive-interaction contract between carbon-based humans and silicon-based intelligences. It transforms exponential full-space-search tasks into linear comparative validation under meta-level axioms, driving down runtime computational overhead and fundamentally dismantling compute-dependence. Large language models function merely as cognitive curtains for linguistic representation; they generate text via data-distribution fitting and possess no inherent ontological constraints, making substrate substitution and paradigm drift endemic. The axiomatic paradigm adopts a cost-structure characterised by upfront cognitive investment during the construction phase and lightweight reuse at runtime. Humans undertake axiom formulation, boundary discernment and ontological exploration; silicon-based agents carry out consistency validation and deductive extension within the axiomatic framework. As a complete meta-benchmark within the publicly-accessible cognitive domain, the Substance-Function-Telos meta-framework is capable of accommodating, examining and adjudicating multiple ontological paradigms (Convergence, Unfolding, Circulation). It simultaneously resolves five inherent pathologies of convergent paradigms: compute dependence, cognitive closure, AI-alignment crises, stifled creativity, and civilisational-heritage alienation. By delineating the applicable boundaries of the axiomatic approach, this paper delivers a self-reflective constructive research programme for AGI alternatives outside convergent paradigms.