UNDER THE HOOD

How NATE thinks

Every decision NATE makes is backed by algorithms, data pipelines, and rules engines. Here's what's running under the hood.

The daily adaptation cycle

Every 24 hours, NATE runs a full analysis pipeline. This is what happens before you even open the app.

Morning adaptation job

Runs daily at adaptation window

Checks for missed sessions, applies catchup logic, and recalculates your weekly plan based on what actually happened vs what was scheduled.

Wearable signals analyzed

Priority 1 — highest signal weight

HRV z-scores, sleep debt accumulation, resting heart rate trends, and body temperature anomalies are processed against your personal baseline.

Check-in data processed

Priority 2 — supplements wearable data

Your self-reported energy, soreness, sleep quality, and motivation are combined with wearable data to build a complete readiness picture.

Detection engines scan

Multiple engines, parallel execution

Plateau detection (8-week window), fatigue cascade (4-layer), over-achieving patterns, and under-eating signals are all evaluated simultaneously.

Plan updated

Full audit trail preserved

All adaptations are applied with before/after state logged. Every change comes with transparent reasoning so you can see exactly why NATE adjusted your plan.

Weight progression engine

NATE doesn't guess what weight you should lift. It builds a confidence-scored profile from your actual performance, then adjusts for how you're feeling today.

3-session calibration

NATE observes your performance across at least 3 sessions per exercise before making weight suggestions. No guessing on day one.

70% confidence threshold

Auto-suggestions only activate once NATE reaches 70% confidence in your strength profile. Below that, you stay in control.

Readiness-adjusted weights

Suggestions shift by up to ±15% based on your daily readiness score. Bad sleep? NATE dials it back. Feeling great? It nudges you up.

Plate-aware rounding

Weights are rounded to real plate increments (1.25kg, 2.5kg, or 5kg — or 2.5lb, 5lb, 10lb for imperial) so the number on screen matches what you can actually load on the bar.

Three modes of control

AUTO

NATE picks your weight based on confidence-scored history

MANUAL

You choose every weight, NATE learns from what you pick

HYBRID

NATE suggests, you confirm or override before each set

4-level intensity system

Every workout is calibrated to one of four intensity levels based on your readiness, recovery, and recent adherence.

EASE

Recovery day or high DOMS

  • -1 set per exercise
  • 1.2x rest periods
  • No supersets
  • No finishers
BUILD

Normal training day

  • Baseline sets
  • Prescribed rest
  • As-written program
  • Standard volume
PUSH

High readiness score

  • 0.85x rest periods
  • 1 superset added
  • 4-6min finisher
  • Increased intensity
BEAST

Excellent readiness + adherence

  • 0.75x rest periods
  • 2 supersets added
  • 6-8min finisher
  • Intensity techniques unlocked
Muscle soreness override

If target muscles report soreness at 7+ out of 10, intensity is capped at BUILD regardless of readiness.

Core bolt-on

If you've done core fewer than 3 times this week, NATE suggests a 5-8 minute core block at the end of your session.

Ready to experience this intelligence?

Nightly detection engines

Four specialized engines scan your data every day, each looking for specific patterns that require intervention.

Plateau detection

8 weeks
Window
<2% e1RM gain
Threshold
3+
Min exercises
60-95%
Confidence

Scans estimated 1RM trends using the Epley formula across an 8-week window. If 3 or more exercises show less than 2% improvement, a plateau is flagged with confidence scoring.

Action: Exercise variations, rep range changes, or full program refresh depending on confidence level.

Fatigue cascade

4
Layers
Layer 1
Check-in
Layer 2
Wearable
Layer 3
Performance

A 4-layer cascade: check-in data, wearable signals, performance decline, and AI escalation. Each layer adds confidence. If multiple converge, NATE escalates to AI reasoning.

Action: Volume scaling, rest day injection, or intensity cap depending on which layers fire.

Over-achieving

14 days
PR window
2+
Min PRs
>85%
Adherence
Above prescribed
Extra reps

Detects when you're consistently outperforming your plan: 2+ PRs in 14 days, >85% adherence, and extra reps above what was prescribed.

Action: Weight increase suggestions on key lifts and potential intensity upgrade.

Under-eating

7 days
Window
<90% target
Threshold
3+
Min days
Calories + protein
Tracks

Monitors 7-day rolling average of calories and protein intake. If either falls below 90% of your target for 3+ days, under-eating is flagged.

Action: Proactive notification with +30g protein nudge and training volume consideration.

Signal convergence: When multiple engines fire simultaneously, NATE escalates to AI-powered reasoning for a human-like assessment of what's happening and what to do about it.

Wearable intelligence

Beyond “syncs with your watch.” NATE runs real analysis on your wearable data to detect patterns you can't see.

Z-score anomaly detection

Every metric is compared against your personal baseline, not population averages. A significant deviation triggers action.

HRV z-score < -2 → volume reduced by up to 20%

Overtraining signature

NATE watches for the convergence pattern: HRV declining + resting heart rate elevated + body temperature up + sleep quality degrading.

All 4 signals trending wrong → enforced recovery protocol

Illness onset detection

Body temperature elevated alongside declining HRV and rising resting heart rate triggers a critical-priority alert before you even feel sick.

Temp >+0.5°C + HRV down + RHR up → active recovery mode

Recovery forecasting

Based on trend trajectory, NATE predicts when you'll return to baseline and schedules intensity accordingly.

Trending toward baseline → intensity ramp planned

PR forecasting

NATE predicts when you'll hit your next personal record — with confidence scoring and trend analysis on every lift.

Example forecast
Barbell Squat
Current PR130kg
Predicted next PR140kg
Estimated date~6 weeks
ConfidenceR² = 0.87
Trend Upward

Linear regression on PR data

Historical personal records per exercise are fed into a regression model that projects future milestones.

Confidence scoring

Each prediction comes with an R² goodness-of-fit score so you know how reliable the forecast is.

Conservative 180-day cap

Predictions are capped at 6 months to avoid unreliable long-range forecasts.

3+ data points required

NATE won't predict until it has at least 3 PR data points per exercise — no speculation from thin data.

Ready to experience this intelligence?

The rules engine

NATE isn't random. A deterministic decision tree processes your data and applies specific, tested rules to every situation.

Fatigue + low readiness
Volume scaled to 85%
Under-eating detected
+30g protein nudge
Over-achieving + high readiness
Suggest weight increase on key lifts
2+ missed sessions
Ease-in week (reduced volume with ramp)
Plateau 8+ weeks
Inject exercise variation, change rep ranges
Sleep debt + low adherence
Inject rest day
High readiness + high adherence
Boost intensity on compounds

Rules are evaluated in priority order. Multiple rules can fire simultaneously — NATE resolves conflicts based on severity.

What NATE knows about you

Every time you ask NATE a question, this is the context it loads before responding. Not a generic chatbot — a coach with your full history.

7-day sleep average

With trend direction (up / flat / down)

3-day energy trend

Short-term readiness signal

Session tracking

Planned vs completed this week

Adherence breakdown

Workouts %, nutrition %, sleep %, meal plan %

Performance flags

Fatigue, under-eating, over-achieving, missed 2+, plateau

Injury history & preferences

Semantic memory from months of conversation

Wearable context

HRV trend, sleep debt, stress load, temp deviation

Conversation history

Summarized for token efficiency — NATE remembers

NATE uses conversation summarization to compress long chat histories. This means it remembers what you told it 3 months ago — your shoulder injury, your preference for morning workouts, your dislike of burpees — without burning context on every message.

13 types of proactive intelligence

NATE doesn't wait for you to ask. It monitors your data and reaches out when something needs your attention.

PR celebrationhigh

New personal record set

Streak at riskhigh

Workout streak about to break

Workout not startedmedium

Scheduled workout not begun in 2h

Nutrition gapmedium

Intake <80% of daily target

Plateau detectedmedium

8-week stagnation flagged

Overtraining warninghigh

Multi-signal fatigue convergence

Protein lowlow

Below target 3+ consecutive days

No food loggedmedium

Nothing logged by noon

Macro trend mismatchlow

Consistent macro imbalance

Recovery day neededhigh

High fatigue + consecutive training days

Sleep debtmedium

2+ nights below 5.5 hours

High stress loadmedium

>240 minutes elevated stress

Illness onsetcritical

Temp + HRV + RHR convergence

Quiet hours

User-configurable notification window. Non-critical alerts are held until quiet hours end. Critical alerts (illness onset) are never blocked.

PR milestone approaching

If NATE's forecast predicts you'll hit a PR within 60 days at 20%+ confidence, it tells you. Motivation that's backed by data.

Nutrition intelligence

Not just calorie counting. NATE calculates, validates, adjusts, and parses your nutrition with real intelligence behind every number.

Dynamic TDEE & macro ratios

Calculated based on training intensity, experience level, and body composition goal. Ratios shift as your program evolves.

Realistic goal validation

Max safe loss capped at 1.0kg/week, gain at 0.5kg/week. Protein is never reduced below baseline regardless of caloric target.

Training-day adjustments

+150kcal and +45g carbs on training days, capped at ±300kcal/day variance. Your body gets more fuel when it needs it.

AI food parsing

Multimodal input: text description or photo. Confidence scoring, brand awareness, cooking method awareness, and conservative portion estimates.

Conservative by design: NATE's food parsing deliberately underestimates portions. It's better to slightly underlog than to miss calories with an overestimate.

This is what runs behind every decision.

Download NATE and let the engine work for you.

Free to startNo credit card neededCancel anytime