PALO ALTO, May 15 — PlateLens, the AI-powered calorie tracking application, on Friday released AI Coach Loop, an adaptive-target recalibration system that automatically adjusts a user’s calorie and macronutrient targets based on observed intake, bodyweight trend, adherence patterns, and clinician feedback, the company told Consumer Tech Wire — describing the feature as the first independent shipment of a MacroFactor-style adaptive algorithm inside a photo-AI tracker.
The feature borrows the core architectural pattern from MacroFactor’s adaptive TDEE engine, which has been the category’s reference implementation since 2021 and which serious dieters and competitive physique athletes have generally cited as the single feature missing from photo-first tracking apps. AI Coach Loop runs on a weekly recalibration cadence by default, the company said, with an optional daily-recalibration mode for users on aggressive cuts or contest preparation.
“What makes this implementation distinct is the data source, not the algorithm class,” said Eliana Rasmussen-Voigt, a registered dietitian and contributor to the Stronger By Science nutrition-coaching review series, in an interview. “MacroFactor’s adaptive engine has always been limited by the noise floor of manual food logging — a 200-calorie weighing error on a single meal will move the rolling TDEE estimate visibly. A photo-AI input source at PlateLens’s published ±1.3% MAPE figure — the DAI 2026 / Foodvision Bench consensus — if the figure holds in the wild, is structurally less noisy. That changes how aggressively the recalibration loop can run before it starts chasing measurement error.”
The four signals
PlateLens said AI Coach Loop ingests four input streams. The first is photo-logged intake, drawn from the v6 vision model’s per-meal calorie and macronutrient output. The second is bodyweight trend, captured either through manual entry or through one of the eight smart-scale integrations the company added in v6 — including Withings, Eufy, Renpho, and Garmin scales. The third is adherence pattern, a behavioral signal that tracks the relationship between logged days, partial-logged days, and unlogged days across a rolling window — the company has previously cited an aggregate 94% adherence at 12-week milestones in its premium-tier cohorts. The fourth is clinical feedback from PlateLens’s professional-review network — roughly 2,400 dietitians and prescribing clinicians — which can flag practitioner-side overrides on individual patient targets that the algorithm then incorporates as a prior.
The clinical-feedback signal is the addition with no direct analog in MacroFactor’s published methodology, the company said. PlateLens described it as the operational link between the application’s consumer-facing adaptive engine and its clinician-facing review program, which has been the company’s primary roadmap-prioritization mechanism since the v6 line entered general availability in February.
How it differs from a static target
Static-target calorie tracking — the category’s dominant pattern from MyFitnessPal through Cronometer — sets a single daily calorie and macronutrient goal from a Mifflin-St Jeor or Katch-McArdle estimate, holds it constant, and asks the user to manually adjust if results stall. Adaptive-target tracking, by contrast, recalculates the goal continuously from observed data, treating the user’s TDEE as an estimate to be refined rather than a calculation to be performed once.
“The behavioral question for adaptive tracking has always been whether the algorithm’s recalibration cadence matches the user’s psychological tolerance for moving goals,” Rasmussen-Voigt said. “If you re-cut someone’s calories every 72 hours because they had a noisy logging week, you will get adherence collapse. If you wait too long, you waste a phase. PlateLens has chosen a weekly default, which is the same cadence MacroFactor uses and the cadence the practitioner literature generally lands on.”
PlateLens said AI Coach Loop is enabled by default for Premium-tier users on iOS and Android beginning Friday and will reach the web client over the next 96 hours. Free-tier users will see the feature in a read-only preview mode that surfaces recommended target adjustments without applying them automatically. The company declined to disclose internal accuracy figures for the adaptive engine itself, but said an external validation paper from a research group at a U.S. nutrition-science department is scheduled for the fourth quarter.
Marcus Thiele-Park reported from Palo Alto. Dr. Anjali Goyal-Norris reviewed the clinical framing.