Program 03 / BioAI

A cross-scale discovery architecture for biology.

BioAI connects Renaissance Field Lite's SQL-backed Mirror Architecture, Universal Data Pattern, and Nesting 1–5 map with B.A.S.I.S. signal intelligence and veterinary-first Quanta Vision. Measurements, transformations, quality, missingness, uncertainty, and support relationships stay attached through the AI process without erasing each field's own instruments, models, or proof rules.

Cross-scale discovery architecture Nesting 1–5 B.A.S.I.S. + Quanta Vision NSF Project Pitch submitted / review pending research stage

Complete program arc

One architecture moves from recurring structure to evidence-bound research products.

01 Mirror Architecture

The SQL-backed relational state-and-evidence system preserves observations, state changes, relationships, transformations, quality, missingness, uncertainty, support status, and memory.

02 Universal Data Pattern

Prior cross-disciplinary mapping exposed a recurring state path that RFL uses to compare structure without pretending different fields share the same measurements or validation rules.

03 Nesting 1–5

The pattern becomes a growing multidisciplinary map across formal systems, matter and chemistry, fields and physical cycles, biology and physiology, and convergence with evidence memory and product routing.

04 Research products

B.A.S.I.S. supplies configurable multimodal signal intelligence. Quanta Vision supplies veterinary-first image, tissue-timeline, retrieval, and review workflows. Both retain field-specific evidence boundaries.

05 BioAI + Nesting Companion

BioAI turns the shared architecture into a research product; the Nesting Companion carries the cross-scale relation map so new biological lanes can be added only as their evidence earns support.

06 Bounded validation to expansion

The NSF Phase I concept tests one biological deployment. A pass establishes a reusable implementation path for later domain-specific work in medicine, materials, physical systems, quantum research, and other mapped fields.

Nesting 1–5

Five connected lanes. Each keeps its own measurements, tools, and validation rules.

The Nesting map gives BioAI cross-scale context without flattening disciplines into one schema. In the proposed biological tests, the five functions govern leakage checks, context, evidence decisions, and routing while samples, labels, and models remain lane-specific.

N1 Formal systems

Identity, transformation, timing, controls, nulls, and the logical rules that determine whether a relation may enter the supported state.

N2 Matter + chemistry

Material, molecular, morphology, composition, and laboratory observations retain their native source and measurement boundaries.

N3 Fields + physical cycles

Energy, environment, waves, periodicity, and physical context become explicit state inputs rather than detached background metadata.

N4 Biology + physiology

Biosignals, images, cells, tissue, interventions, and outcomes are organized by scale, time, quality, recurrence, and evidentiary role.

N5 Convergence + evidence memory

Supported relations converge into traceable memory, reviewer decisions, failure routes, and research-product workflows without promoting open hypotheses into facts.

Integrated research surface

BioAI is larger than either product lane.

Field-native evidence Typed biological state Nesting context Model + evidence decision Research product
Product 01 / B.A.S.I.S.

Configurable signal intelligence

Physiological, behavioral, environmental, laboratory, and operational signals enter through adapters that preserve units, clocks, quality, artifacts, continuity, and source identity. HRV and EEG are current routes, not the product limit.

Product 02 / Quanta Vision

Veterinary-first visual intelligence

Images, lesions, wounds, tissue surfaces, timelines, retrieval, and veterinarian review form a separate visual lane. Human medical imaging remains a future path requiring its own data, validation, partners, and regulatory review.

Expansion map / Nesting Companion

Add depth without erasing boundaries

The companion map is the route for cellular, molecular, imaging, environmental, assay, and intervention lanes. Each enters only with its own adapter, evidence contract, controls, and release gate.

Research product / BioAI

Operate across the lanes

BioAI compiles the state, runs bounded learning and baseline comparisons, tests degradation and uncertainty, and returns every result with source-resolved evidence for researcher review.

Product architecture

Five modules can run together or around an existing research stack.

01 / Data adapters

Bring the sources in

Biosignals, imaging, morphology, molecular, environmental, intervention, and outcome data enter with source identity and quality intact.

02 / State compiler

Build the typed graph

The compiler and validator preserve measured, derived, inferred, imputed, missing, and excluded states alongside transforms and permitted relationships.

03 / State operator

Keep comparison honest

Customer models, modality-specific models, classical baselines, flat fusion, typed graph operators, and structure ablations run through one bounded comparison harness.

04 / Evidence router

Expose what holds

Quality, degradation, missingness, calibration, uncertainty, exclusions, unsupported relations, and failure routes remain visible in every evidence packet.

05 / Research console

Make the state reviewable

Researchers inspect the source path, model output, active relationships, uncertainty, missing evidence, exclusions, and next validation gate in one surface.

Current component support

The integrated hypothesis is open. The underlying state, capture, routing, and control work is real.

Measured physiology

Repeated state capture

Fifteen valid repeated condition rows carried target, calm-control, and drift-control states across 98,027 Muse packet rows, 9,750,684 decoded samples, and 3,980 HRV observations. The strongest current support is HRV separation; other lanes remain quality bounded.

Runtime routing

State and continuity stay visible

RFL normalized 468 recorded state-frame rows and reconciled 480 expected source, guard, and sink events across eight named lane sinks with zero observed drops and zero sequence gaps.

Real-data controls

Different biological data types

Cell-morphology and clinical/metabolic benchmark surfaces were run against shuffled-label and feature-shuffle controls, establishing reusable packet, control, and receipt discipline across distinct biological inputs.

Support boundary

Components, not a pre-claimed outcome

These receipts reduce implementation risk. They do not establish the proposed integrated BioAI advantage, population validation, diagnostic utility, or clinical performance.

Proposed Phase I validation

Two products. Three test surfaces. One shared evidence architecture.

The bounded NSF design tests how the larger architecture behaves in biology. Human physiology and veterinary imagery remain independent populations with separate labels, models, splits, and domain rules. No human–dog correspondence or sample-level fusion is assumed.

Lane A1 / B.A.S.I.S.

Protocol-specific state transitions

PhysioNet stress and structured-exercise data test stress/rest, aerobic baseline-work-cooldown-recovery, and anaerobic sprint-recovery tasks without inventing cross-protocol subject continuity.

Lane A2 / B.A.S.I.S.

Degradation + redundant evidence

ScientISST MOVE tests masking, corruption, motion, missing channels, and alternate routing while paired sensors remain source-distinct rather than interchangeable ground truth.

Lane B / Quanta Vision

Veterinary image robustness

A permissively licensed 95-dog dataset uses dog-grouped splits for classification, image degradation, calibration, and evidence-trace tests inside the veterinary boundary.

Shared architecture test

Nesting functions without false fusion

Across all three evaluations, Nests 1–5 govern leakage checks, context, evidence decisions, and routing while samples, labels, models, and field rules remain separate.

Prospective gates

Clean-data noninferiority; at least 80% retention under the primary degradation condition; calibrated uncertainty; 100% output-to-source traceability; zero silent imputation; unsupported relations fail closed; and reproducible evidence manifests. Predictive superiority remains a stretch target.

Multidisciplinary expansion

Biology is the first bounded wedge, not the edge of the system.

Biological depth

More modes, earned one lane at a time

Cellular, molecular, omics, assay, imaging, environmental, and intervention routes can join the architecture after their datasets, relations, controls, and validation boundaries are frozen.

Research teams

Connect what already exists

Biotechnology, biopharma, contract-research, digital-biomarker, translational-laboratory, academic-core, veterinary, and government teams can retain their instruments, models, storage, and domain experts.

Deployment

Full operator or protected state layer

Teams can use the complete BioAI research operator or connect the state-and-evidence layer to their own models in private, local, or customer-controlled environments.

Beyond biology

Reusable architecture, separate proof

Medicine, materials, physical systems, quantum research, and other mapped fields are continuation lanes. Each requires its own domain-specific evidence before public support claims advance.

Evidence boundary

Current support: a cross-scale state-and-evidence architecture with working product components.

RFL has implemented the SQL/SQLite state and memory layer, a growing Nesting 1–5 map, B.A.S.I.S. capture and routing surfaces, BioAI research and operator surfaces, and a veterinary-first Quanta Vision prototype. The NSF Project Pitch was submitted on October 1, 2026 and is awaiting review. The integrated comparison, every proposed performance gate, and generalization across the complete Nest map remain validation work—not prior results.

Next gate

Freeze the eligible datasets and evidence contract, complete the compiler and validator, execute all three independent test surfaces, and return every result as a source-resolved evidence card. Clinical, diagnostic, human-imaging, and additional cross-domain releases remain separate evidence and regulatory gates.