How Inner Navigation works, under the hood.
Everything on this page is public for a reason. If you’re going to trust your data to a project, you should know exactly how it’s collected, protected, and analyzed.
Where the external data comes from.
Inner Navigation does not generate cosmic or geophysical data. It pulls from public, peer-reviewed, or officially maintained scientific sources. Every data stream is attributed and timestamped. The stack is listed below in full — if a source isn't on this page, Inner Navigation isn't using it.
- [01]
Local weather
Current conditions and historical backfill over the last 90 days. Temperature, precipitation, wind, UV index, atmospheric pressure, solar radiation, sunrise and sunset times.
SourceOpen-Meteo (forecast and archive endpoints)
- [02]
Air quality and pollen
Airborne pollen counts (grasses, birch, olive, ragweed and others) and the European Air Quality Index from Open-Meteo's dedicated air quality service, combined with ground-truth measurements (PM2.5, PM10, NO₂) from the nearest physical monitoring stations.
SourceOpen-Meteo Air Quality API and OpenAQ
- [03]
Space weather
Planetary K-index (Kp), F10.7 cm radio flux, real-time solar wind parameters, interplanetary magnetic field (including the Bz component), GOES X-ray flux. Refreshed daily at 06:00 UTC.
SourceNOAA Space Weather Prediction Center (SWPC)
- [04]
Solar and geomagnetic events
Solar flares, coronal mass ejections, geomagnetic storms, and solar energetic particle events. Refreshed daily at 06:00 UTC.
SourceNASA DONKI (Space Weather Database Of Notifications, Knowledge, Information)
- [05]
Seismic activity
Earthquake events and magnitudes, queried two ways: locally (filtered by radius around your approximate position) for your personal dashboard, and globally (via scheduled ingestion) for the community-wide event log.
SourceUSGS Earthquake Hazards Program
- [06]
Lunar and planetary positions and aspects
Heliocentric and geocentric ecliptic longitudes, along with active aspects between bodies. Computed locally using the Swiss Ephemeris engine, which derives from NASA JPL ephemerides. No external API call required — ephemeris calculations run on our own servers.
SourceSwiss Ephemeris · NASA JPL ephemerides (local computation)
On aspect calculation
Aspects are evaluated between all major bodies (Sun through Pluto, plus the lunar nodes). Major aspects — conjunction, sextile, square, trine, opposition — are active by default. Minor aspects — semi-sextile and quincunx — can be enabled for users who want a finer-grained view. Standard orb tolerances apply (8° for conjunctions and oppositions, 6° for squares and trines, 4° for sextiles, 2° for minor aspects), with a +2° bonus when the Moon is involved, consistent with traditional astrological practice. Orb values are user-configurable for researchers who want to test tighter or looser thresholds.
Architecture
Weather, air quality, and pollen data are queried directly from the client device using your approximate location. Space weather, seismic, and astrological data pass through a dedicated worker that aggregates, caches, and normalizes the upstream sources into consistent daily time series. This keeps load off the public agencies we depend on, keeps response times fast for you, and creates a single auditable point where data quality can be checked.
The inner data layer.
External data is half the picture. The other half is you — logged deliberately, in your own words, using the categories that matter to you.
Core tracked variables
- Mood (multi-dimensional, not a single happy-sad scale)
- Physical energy
- Willpower and focus
- Sleep duration and subjective quality
- Free-form journal notes (typed or dictated)
- Food intake (freeform or structured)
- Exercise intensity and type
- Meditation or breathwork
- Creative and cognitive output
Custom variables
Anything you want. The more honestly you log, the more the correlations surface.
- Menstrual cycle phase
- Work productivity
- Conflict with a specific person
- Dreams
- Caffeine timing
- Social interactions
Logs take roughly two minutes a day. The app nudges you gently — not obsessively — at times you choose.
From logs to patterns.
Correlations in Inner Navigation operate on three levels.
- [01]Personal
Personal patterns
For your own data alone, the app computes rolling correlation coefficients (Pearson and Spearman) between each inner variable and each external variable, across multiple time windows. Results become visible to you once you have roughly three weeks of consistent data — enough for weak signals to begin emerging without being dominated by noise.
- [02]Contextual
Contextual baselines
Your personal patterns are weighted against your own established baselines, not generic population averages. A "low energy day" for you is defined relative to you, not relative to some standardized scale.
- [03]Collective (opt-in)
Collective patterns
Pseudonymized contributions to the community are aggregated using robust statistical methods, with k-anonymity (k ≥ 10 contributors per cell) and ±0.3 noise applied before publication. Outliers are handled with care. Sample size disclosures are always shown alongside any collective finding — if only 42 people logged on a given day, the report says so.
Multi-variable patterns, lagged effects, thresholds.
The optional AI layer uses commercial large language models to help describe multi-variable patterns that linear correlation misses — interactions between three or four variables at once, lagged effects, threshold behaviors. User data is not used to train a general AI model, and the methodology behind this layer is documented separately in our technical notes, linked at the bottom of this page.
How your data is protected.
This section is the one you should read twice.
- [01]
On-device by default
Your raw logs are stored locally on your phone. They are not transmitted to our servers unless you explicitly opt in to a specific connected feature such as personal cloud backup, community statistics, or AI analysis.
- [02]
HTTPS and HSTS enforcement
The .app TLD enforces HTTPS connections at the DNS level, meaning your connection to our servers is encrypted before the browser even loads. No insecure fallback is possible.
- [03]
Pseudonymization at rest, k-anonymity at publication
When you opt in, direct identifiers are stripped on your device before transmission: no name, no email, no device ID, no GPS coordinates. Location is generalized to a regional bucket (~110 km cell). Stored data is pseudonymized in the GDPR sense (Article 4(5)) — still linked to your account so we can honor a deletion request. What is then published to the community as collective insights is aggregated with k ≥ 10 contributors per cell and statistical noise (±0.3) — no individual remains attributable in those outputs.
- [04]
Separate opt-ins by purpose
Personal cloud backup, community statistics, community location, AI analysis, text notes for AI, and personal AI memory are separate choices. None of them is enabled by accepting another one, and withdrawing one stops that specific flow.
- [05]
Full export, full delete
One tap gives you your dataset as an export file. Account deletion removes hosted personal data through the app or web account flow, subject to narrow legal retention where required.
- [06]
No third-party sale. Ever.
This isn’t a marketing claim. It’s a constitutional commitment of the project. Your data is never sold, never rented, never shared with advertisers, never monetized through any channel other than the optional paid extras — AI analysis and cloud sync — that you pay for directly.
What correlations mean, and what they don’t.
A correlation is not a cause.
The single most important sentence on this page.
If Inner Navigation shows that your sleep quality drops during geomagnetic storms, it does not prove that geomagnetic storms cause your sleep to suffer. It means: over the period observed, these two variables moved together. Possible explanations include direct causation, reverse causation, a common third cause, or coincidence. The app will show you the correlation. The interpretation is yours.
- 01Sample size matters
Weak correlations in small datasets are often noise. The app displays confidence indicators on every finding, and hides findings below a minimum threshold of reliability.
- 02Multiple comparisons are disclosed
When you run correlations across dozens of variable pairs, some will light up by chance alone. The methodology accounts for this through statistical corrections, and the corrected results are what you see.
- 03We publish failures, not just successes
In quarterly Collective Reports, findings that did not replicate or that weakened as sample size grew are reported alongside those that strengthened. A project claiming only positive results is a project hiding something.
What Inner Navigation is not designed to do.
- [01]
Not a medical tool
Nothing in the app constitutes medical advice, diagnosis, or treatment recommendation. If you observe a concerning pattern — persistent low mood, sleep disturbance, energy collapse — please consult a qualified professional.
- [02]
Not a predictive oracle
We do not tell you what tomorrow will feel like. We help you understand the forces that may correlate with how you feel, over time, so you can make better-informed decisions. The difference is important.
- [03]
Not a substitute for self-knowledge
The app surfaces patterns. It does not replace the attention, honesty, and curiosity you bring to your own life. It is an instrument, not an authority.
- [04]
Not infallible
The external data sources we draw from occasionally have gaps or errors. The correlations we surface carry statistical uncertainty. The AI analysis layer is a tool, not a truth machine. Treat every finding as a hypothesis worth testing against your own experience.
Where this is going.
Inner Navigation is a long-term research project, not a launch-and-forget product. The research programme only starts in earnest once the community passes roughly 1,000 active contributors — below that, aggregate findings wouldn't be statistically honest. What follows is the intended path, not a guarantee.
- [Y1]Year one
Seed the community
Grow a committed contributor base. Once it passes roughly 1,000 active contributors — the point where k-anonymous aggregates (k ≥ 10 per cell) become statistically meaningful — the first Collective Reports begin. Until then, the focus is the pseudonymization and k-anonymous aggregation pipeline and the AI analysis layer, refined on community feedback.
- [Y2]Year two
Open the methodology
Release k-anonymous aggregate datasets (k ≥ 10, with statistical noise) to independent researchers on request, under clear ethical frameworks. Publish the statistical methodology and code of the correlation engine in open source.
- [Y3+]Year three and beyond
Open the methods, publish the findings
Open-source the statistical methodology, aggregate-data schemas, and analysis tooling as the community reaches sufficient scale. If the project gets that far, we will seek to work with established researchers in chronobiology, heliobiology and related fields, and to publish findings in peer-reviewed or peer-review-adjacent venues.
The knowledge this community generates belongs to the community. Inner Navigation is a steward, not an owner.
Go deeper.
Full methodology
Correlation engine architecture, statistical methods, AI layer model cards, pseudonymization and k-anonymity specifications.
[linked once ready]
Researchers & institutions
Academic institutions and aligned projects welcome.
research@innernavigation.appChallenge us
If we got something wrong, we want to know.
support@manifestwealth.ioTransparency is non-negotiable · Methodology is public · Findings will be open-sourced