Known Limitations

Status: local-alpha draft.

Product Limitations

MindCanary is not a diagnostic, prediction, emergency, or treatment product.
Insights are descriptive comparisons against local records, not clinical interpretations.
The app may be useful before baseline history exists as a logbook, but pattern explanations require enough comparable data.
Missing days remain missing. They are not treated as zero.
Baseline descriptions currently pool all prior days containing each signal. User-defined work/non-work comparison calendars are not implemented. A sustained schedule-related change may therefore be described without MindCanary claiming what caused it; supporting dates remain visible.
Switching and active-time descriptions normalize by recorded coverage. They describe activity during observed periods, not whole-day behavior when a connector was absent.
Tracking can increase rumination for some people. Pause and deletion controls are part of the product, not edge cases.

The first supported packaged target is Linux on Pop!_OS/Ubuntu-like systems.
GNOME/X11 active/idle duration is the first OS adapter.
Wayland, KDE, macOS, Windows, mobile, and foreground-app categories require separate adapters and review.
Lock/unlock and suspend/resume counts require the GNOME ScreenSaver and logind event streams. They remain unavailable when those local services do not connect.
The browser extension is optional. The desktop app remains useful with check-ins only.

Alpha Chrome setup uses a bundled unpacked extension until the Chrome Web Store listing is ready. Testers must reload or replace that unpacked folder when a new alpha bundle ships.
The extension stores only aggregate batches and operational queue state, not URLs, page titles, page text, search terms, or browsing history.
Browser aggregates are delivered in aligned 15-minute periods, so the desktop may not update immediately after installation.
Disabling or removing the extension is controlled by the browser. MindCanary can show that browser data has not arrived recently, but it cannot manage browser-owned extension storage from the desktop app.
The extension retry queue is bounded. If the daemon is unavailable for long enough, old unsent batches are dropped and counted rather than retained indefinitely.
Firefox builds and native-host manifests are implemented, but temporary add-on installation and native-message delivery still need real-session validation. Chrome remains the first packaged connector.
Use only one browser connector at a time. Records do not yet carry connector provenance, so overlapping Chrome and Firefox collection would double-count shared activity.
Continuous-scrolling collection is an optional aggregate-only adapter for x.com and twitter.com; it is not a general website tracker.

Clearing records inside the app is not the same as uninstalling MindCanary.
Destroying the local database profile removes database files and the OS-keyring database key, but it does not remove browser extension storage or user-created exports/backups.
Repeated same-day check-ins remain separate in encrypted storage, while History and the current CSV export show their daily average and count. A moment-by-moment check-in view and export are not implemented yet.
Export files are plaintext readable files in a folder the user chooses.
Losing local records before exporting or backing them up cannot be reversed by MindCanary.
An encrypted backup cannot be restored without its generated recovery secret. MindCanary does not retain a recovery copy.

Local encryption protects copied database files and powered-off devices better than plaintext storage, but it cannot protect against malware running as the same unlocked OS user.
A malicious authorized extension update can observe whatever browser permissions allow. Permission minimization and review reduce this risk but do not remove it.
Hosted sync, hosted dashboards, payments, telemetry, and AI are deferred until they have separate security and privacy designs.