1. Problem
Nutrition care suffers a structural data gap. Between clinical visits, the data flowing back to the dietitian is either self-reported (food journals, recall surveys) or absent. Self-report has well-characterized limitations: under-reporting, recency bias, social desirability, and the cognitive load of logging. Absence of data, in turn, forces practitioners to make protocol decisions from snapshots taken weeks or months apart — long after the moment of intervention has passed.
The cost of this gap is measurable. Attrition from private-practice nutrition care exceeds seventy percent within the first three months. Patients report feeling unsupported between visits; dietitians report difficulty driving behavior change in the absence of granular signal. Existing CGMs, food-photo apps, and behavioral logging tools address fragments of the problem but introduce their own friction: glucose without the food, food without the context, context without the clinician.
2. Approach
Spoonified begins at the utensil. A camera, haptic, and audio-visual sensor array embedded in the silverware captures the transit between plate and mouth — what is being eaten, the cadence of consumption, the duration of the meal, the context of where and with whom — without requiring active user input. Edge inference identifies food composition and behavioral signals on-device; structured data is uploaded asynchronously to the patient's care record.
The dietitian sees this data as a structured timeline in a dashboard purpose-built for between-visit care. Alerts are generated by interaction rules (drug–food windows, glycemic load thresholds, ultra-processed flags). AI-assisted message drafts are surfaced to the dietitian for review and edit; the patient never receives automated clinical advice directly.
3. Novel measurements
Beyond replicating the data captured by existing food-log tools, Spoonified introduces two novel indices currently under validation:
· A satiety index, modeled from bite cadence, meal duration, glycemic load, and post-meal hunger return. · A satisfaction index, modeled from a brief post-meal self-report paired with behavioral indicators.
These indices are intended to complement, not replace, established subjective scales (e.g., visual analog hunger and satisfaction scales). Calibration studies pairing Spoonified output against gold-standard subjective measures and continuous glucose monitoring data are in active design.
4. Architecture
The platform is structured around a source-agnostic data model. Every captured record — whether from Spoonified silverware, a connected CGM, a wearable, or a lab integration — carries the same shape. New data sources hydrate the existing dashboard surfaces without UI changes. This separation of capture, schema, and surface is intended to support future deployment across clinical, longevity, professional sports, and consumer contexts without architectural divergence.
AI is positioned as a force-multiplier for the clinician, not a replacement. The dietitian remains the clinician of record; the patient does not receive automated clinical advice. All AI-generated content is labeled when surfaced to either party.
5. Validation roadmap
Three studies are planned in sequence:
· An outcome-cohort study comparing HbA1c, HOMA-IR, ApoB, fullness, and satisfaction trajectories across patients using Spoonified versus standard nutrition care. · A satiety-index calibration against validated subjective hunger scales (Flint, Blundell) in a controlled feeding-room setting. · A satisfaction-index construct-validity study distinguishing satisfaction from hedonic liking and wanting.
Each study will publish methodology, results, and limitations. IRB review and informed consent are in scope as a precondition to data collection from any non-demo user.
6. Compliance and ethics
The platform is being built to be HIPAA-compliant when patient data is connected. AI suggestions are reviewed and sent by the dietitian; patients do not receive automated clinical advice. Data is portable: patients can export everything captured about them, at any time. Sale of patient data to third parties is not on the roadmap.
7. Open questions
This document is a starting point, not a final position. Areas where outside expertise is welcome:
· The optimal weighting of behavioral, metabolic, and self-report signals in the satiety index. · The most defensible operational definition of satisfaction in a clinical context. · The ethical framework for AI-suggested message drafts in nutrition care. · The path to FDA clearance, if any, for the satiety and satisfaction indices.
Comments are encouraged via the feedback chat inside the demo, or in correspondence with the team.