How we build Omnio — health data infrastructure, MCP servers, time-series analytics, and the AI layer that ties it together.
Energy availability, RED-S risk factors, biomarker signatures of underfueling, and the confounders that mimic them — plus where data stops and clinical assessment begins.
Consumer EEG is noisy. Muscle activity, blink, jaw clench, and electrode liftoff produce spectral signatures that drown the real brain signal. How autoencoder-based artifact detection separates brain from noise.
Mediterranean, ketogenic, high-protein, plant-based, high-processed, balanced, low-carb, high-carb. Weekly logs reveal which pattern you actually follow — often different from what you think.
Bench, squat, and deadlift progress at different rates for structural reasons. A one-size-fits-all progression rule under- or over-loads every lift.
Two devices measuring the same night and reading 58 and 79 is telling you something — usually that the composite score shouldn't be a single number at all. How cross-source validation handles divergence.
Per-food GI tables ignore that nobody eats a single food alone. Protein, fat, fiber, and cooking method all shift the curve. Meal-level glycemic load is closer to what your glucose actually does.
Self-reported RPE drifts with fatigue, stress, and experience. Calibration against objective markers (velocity, RIR, recovery) catches the drift.
When you eat changes outcomes independent of what you eat. Eating window, meal spacing, and circadian alignment are measurable, and they matter. Here's how to track them.
Your primary wearable didn't sync last night. Your dashboard shouldn't just go blank. Source fallback chains degrade gracefully across devices while preserving confidence.
Your chest isn't recovered in the same window as your quads. Here's how per-muscle recovery half-lives change training frequency and session design.
Total polyphenols in milligrams misses the point. A diet hitting six polyphenol classes outperforms one hitting a single class at higher dose. Here's how class diversity tracking works.
Strength is more durable than aerobic fitness. Here's what the detraining literature shows about what you lose, when, and how to ramp back safely.
A readiness score with no confidence value is an opinion, not a measurement. How to read credible intervals, spot low-confidence days, and know when to suppress.
NOVA sorts food by processing, not calories. A 400-calorie bowl of oats and a 400-calorie bowl of sugary cereal differ in ways macros cannot capture. Here's how the four groups work.
MEV and MRV are population ranges. MAV is the per-muscle weekly ceiling your body can currently tolerate — learned from your history, not a table.
NOVA processing, polyphenol diversity, meal-level glycemic index, chrono-nutrition, IARC exposure, and 35 micronutrients — the nutritional dimensions calorie counters ignore.
How per-muscle volume tolerance, recovery half-lives, and Bayesian load models translate raw wearable data into actionable training prescriptions.
Composite health scores fuse many inputs into one number — but only if you know which inputs are trustworthy. Confidence, cross-validation, suppression.
Your bathroom scale's body fat number is a guess with a ±5% error margin. Here's what DEXA, bioimpedance, and calipers actually measure, how accurate each is, and how to use them together.
A 75 on Oura is not a 75 on Garmin. Here's what each sleep score actually measures, which factors predict how you feel, and why consistency matters more than any single number.
Single metrics lie by omission. A composite score synthesizes HRV, sleep, training load, and recovery into one number — but only if you can see how it's built.
The acute-to-chronic workload ratio is the single best predictor of training-related injury. Here's what it measures, where the 0.8-1.3 sweet spot comes from, how Omnio calculates yours, and the mistakes that get people hurt.
Lab reference ranges are built for the general population. If you train seriously, most of your results need a completely different interpretation.
Garmin Body Battery, WHOOP Strain, Oura Stress, WHOOP Recovery — these scores sound like they measure the same thing. They don't. Here's what each one actually calculates, why they contradict each other, and how to read them together.
RMSSD, SDNN, LnRMSSD — your wearable picks one and doesn't tell you which. Here's what HRV actually measures, how Oura, Garmin, WHOOP, and Polar each capture it differently, and why your baseline matters more than any single number.
Bevel is a polished health dashboard with Oura and Garmin support. But if you want validated nutrition data, transparent scoring, or cross-domain correlations backed by statistics — here's what else is out there.
Cronometer is the gold standard for micronutrient tracking. But if you want your nutrition data connected to sleep, HRV, and training — or a modern mobile experience — here are the best alternatives.
Levels made continuous glucose monitoring mainstream. But CGM data alone doesn't tell the whole story. Here are the best alternatives for understanding your metabolic health.
MacroFactor's adaptive algorithm is best-in-class for macro coaching. But if you want micronutrients, meal quality, or wearable-connected insights, here are the alternatives worth considering.
MyFitnessPal counts calories. These alternatives track meal quality, micronutrients, and how your diet affects sleep, recovery, and training. Here's what to switch to.
Your wearable data contains early warning signals that precede readiness dips by 1-5 days. We built a system that learns which signals matter for you personally — and warns you before the dip arrives.
Your wearable data contains a hidden signal: the ultradian rhythm. We're building a system that reads it — predicting your best focus windows from sleep data, then refining in real-time with a heart rate monitor.
Calorie counting is table stakes. Omnio validates your logs against government nutrition databases, scores meal quality, tracks 35 micronutrients and polyphenols, classifies your dietary pattern, and connects what you eat to how you sleep, recover, and train.
When you connect Oura, Whoop, or Withings to Omnio, you sign in on their website, not ours. Here's what OAuth 2 does, what we receive, and what we never see.
Most calorie calculators give you a number based on a formula from the 1990s and call it a day. Omnio watches what you actually eat, how your body actually responds, and learns your real metabolic fingerprint over time.
Your AI health assistant analyzes your sleep, recovery, and training data — but it never sees your name, email, or account ID. Here's how we built privacy into the architecture itself.
We asked Oura, WHOOP, and Omnio the same research question. Two gave vague platitudes with zero citations. One gave verifiable evidence. Here's why the difference is architectural.
Omnio doesn't just track your workouts — it derives your training schedule, gates every session through biometric readiness, adjusts for nutrition, and learns from outcomes. Here's how we're building a genuinely adaptive training system.
We reviewed 17 peer-reviewed studies comparing Oura, Apple Watch, Fitbit, Garmin, WHOOP, and Samsung across sleep, heart rate, HRV, SpO2, VO2 max, and more. Here are the specific numbers.
Every fitness app says it 'learns.' We wanted to prove it. Here's why we chose Bayesian parameter estimation over neural nets, how six independent sub-models personalize your training, and why the system can never be worse than a textbook.
Readiness scores tell you how recovered you are. They don't tell you whether skipping today's session is smart load management or the start of detraining. We added a sports science metric to close that gap.
When you test enough metric pairs, something will appear correlated by chance. Here's how Omnio's correlation engine uses detrending, dual methods, p-values, and FDR correction to separate real signal from noise.
Every major wearable now has an AI assistant. We asked all three to compare a week of sleep data. The difference in depth reveals a fundamental architectural gap.
Most health platforms treat AI as a chatbot skin over a database. We built an MCP server with a plugin architecture, time-series storage, and multi-step reasoning. Here's why.
Using Oura for sleep and Garmin for training? Here's how to combine both data sources into one dashboard without duplicating metrics or losing context.
Garmin and Oura solve different problems. Here's how they compare for sleep tracking, training readiness, workouts, HRV, and whether you should use one or both.
Trying to choose between Oura and WHOOP? Here's the practical difference in sleep tracking, HRV, recovery, workouts, battery life, and who each device is actually best for.
Looking for the best app for a Polar H10 chest strap? Here's what to look for if you care about heart rate, RR intervals, HRV, ECG, guided readiness checks, and long-term trends.