Healthcare is rapidly advancing toward artificial intelligence, which may yield significant positive impacts. For example, an AI model trained on extensive images, records, and waveforms can identify patterns that may elude human observers. However, we have deliberately chosen a different approach, and it is essential to clarify both our rationale and methodology.
This technology does not incorporate any trained or learned AI models. Each result produced by the system is a declared, deterministic function of the measured data, which can be replayed from the raw record and independently verified. This design choice is driven by the specific nature of the measurements and the potential implications of their interpretation.
A state-of-the-art AI model in 2026 operates by example: when exposed to sufficient labeled cases, it identifies patterns that distinguish them. While this capability is powerful, the model's underlying assumptions are typically implicit, as every training set embeds a specific ontology. Data are collected by particular instruments, labeled using defined vocabularies, sampled from specific populations, and categorized according to pre-existing frameworks such as disease labels, coded diagnoses, conventional biomarkers, or charted outcomes. Although AI models can skillfully learn correlations within these parameters, they do not explicitly state the perspective from which their answers are derived. They do not declare their ontological framework and cannot operate beyond it to recognize patterns outside their training.
This inherited ontology becomes critical when the subject of measurement is novel. Boundary exchange represents a new measurement object for which no dataset, prior labeled examples, predicate device, or established category exists. A model trained on existing medical data would necessarily interpret the new measurement through its pre-existing ontology, returning disease labels, biomarkers, or risk classes already present in its training. While the detected pattern may be novel, the interpretive categories remain unchanged. Building upon a learned model in this context would reinforce the existing ontology, which we intentionally avoid for reasons detailed in our previous work.
In contrast, a formal algebra explicitly defines its position prior to making any claims. It specifies the object of analysis—such as a boundary, source, witness, response, residue, or residual—and assigns, for each reading, a grade indicating the status of the claim: admissible or set aside, source-underclosed or physically witnessed, forward-consistent or pending, permitted claim or feature not yet meaningful. Each result transparently documents its progression: from the initial source, through the recorded witness, assessment of admissibility, evaluation of forward consistency, to the final claim grade and stated limitations. Every assertion is fully traceable through this documented process.
This distinction is particularly significant in high-consequence domains. Measurements introduced into medicine must clearly delineate their permissible interpretations. It is essential to distinguish between a signal and a claim, a feature and a diagnosis, a correlation and a measurement, and a measurement and an inference, explicitly indicating the stage reached at each step. While a learned AI model may return a class or score, a formal algebra provides a traceable authorization: indicating what was witnessed, what was admissible, which residues persisted, whether laminarity was maintained or disrupted, whether the far-side account is forward-consistent, whether the claim is provisional and not yet of bodily origin, and whether a feature is not yet suitable for clinical interpretation.
This methodological stance precedes the application of AI and establishes a framework within which AI can operate appropriately. A learned model can identify patterns only within a pre-defined measurement space; the formal algebra delineates this space and determines the reliability of its readings. Once boundary-exchange measurements are standardized and sufficient evidence is accumulated, AI models may play a valuable downstream role as interpreters, discovery engines, or cohort analyzers. When applied after measurement definition, AI analysis can facilitate and accelerate the identification of discoveries, associations, patterns, and potentials. If applied prematurely, however, AI would impose its ontology, causing the new category to revert to established frameworks before it is properly characterized.
Much of healthcare AI development follows a sequence of accumulation: constructing a data lake, training a model, and generating clinical predictions. While this approach may enhance predictions within existing medical categories, our innovation is structured as an explicit measurement sequence: comprising an instrument, a witness chain, a formal algebra, an admissible measurement object, and a claim-qualified interpretation. This methodology establishes the prerequisites for a new category to be articulated without assimilation into pre-existing frameworks. Our custom algebra was not designed as a mere academic embellishment; rather, it provides the structure that translates instrument measurements into an unprecedented perspective on the human body, enabling the integration of living boundaries into medicine while preserving terminological precision. From an investment perspective, this serves as the moat, the safety architecture, the regulatory foundation, and the scientific identity of Fieldflux Biosystems.
On the ontology that motivates this measurement space, see The boundary comes first. On the algebra that makes claim authorization explicit, see A boundary-observable certification algebra. On how a reading is challenged before it earns authority, see How we try to make the signal fail.