Tepna Working Preprints

Findings from the physiological-signal suite, written up so they need not be rediscovered. Each paper regenerates its tables and figures deterministically from a named, local analysis tool. Standardized format: paper.css. Drafts — not peer-reviewed, not for clinical use.

Candidate papers in the backlog (forward agenda — see PAPERS-ROADMAP-2026-06-24-BRIEF.md)

The series is lopsided toward simulation scored by real detectors; the planned stack (vendor-adapter layer + multi-vendor unifier, OverDex, EEGDex, Ultrahuman/Spiro nodes) opens the real-world validation front. Each candidate is tagged with its stack dependency. Honest sim/real labelling and a named regeneration tool remain mandatory.

Buildable now — no new node

  1. Timestamp-pathology benchmark. A deterministic corpus of consumer-export timestamp failure modes (DMY/MDY ambiguity, zoned vs floating, time-only midnight-roll, 14-digit, epoch-vs-civil) with the Clock Contract's floating tMs model stated as a citable method. The narrowest first paper already named by synthetic-data-frontier; corpus mostly exists in the parseTimestamp tests. methods/repro LOW effort · now
  2. The dead-ends paper — negative results as a deliverable. What doesn't survive scrutiny, with evidence: optical PRV ≠ RR/ECG HRV (+32%); Welltory wellness composites collapse to driver effects (and emit a fake 0 when inputs are absent); daily CGM-CV ICC≈0; SQI green while beat-yield fails under apnea; rolling-mean ODI self-suppression. A citable "map of the walls in the maze" — the manifesto made into an artifact. synthesis LOW–MED · now
  3. Cross-signal plausibility as automated QC. Encode physiological coupling laws (apnea→desat→HR surge; exertion→HRV drop; glucose↔HRV) as a forward model, run it backward to flag impossible combinations as artifacts or mis-routed files — QC no single-signal SQI or filename sniffer can do. Also the intelligence layer OverDex wants. sim + real MED · now

Unlocked by the planned stack — the real-validation front

  1. Real-data multi-vendor HRV agreement — the adapter-layer flagship. The real-world sequel to rmssd-equivalence: pool real co-recorded consumer devices across vendors (Polar, Coospo, Wahoo, Garmin…) and test rMSSD/SDNN agreement reference-free via the three-cornered hat. A hat across N consumer vendors is genuinely rare. real data needs adapter Phase 1 · MED
  2. EEG-anchored validation of cardiorespiratory sleep proxies — highest scientific payoff. EEGDex (Muse) supplies a near-PSG sleep-stage reference; ask how well ODI / HRV / movement recover EEG-defined sleep architecture. Upgrades the suite from "synthetic ground truth" toward a validated anchor. real data needs EEGDex · HIGH
  3. Longitudinal reference-free σ drift — OverDex-enabled metrology. sigma-no-reference pinned σ from one ~2 h hat; OverDex auto-detects every co-recording in an archive → track each device's σ over months (does consumer-sensor error drift, caught with no reference?). sensor-trio-nights already did the power analysis. real data needs OverDex · MED
  4. Real-data metabolic-autonomic coupling — UltrahumanDex sequel to cgm-hrv-coupling: the first real test of the synthetically-predicted glucose↔HRV shared-driver effect on co-worn data. real data needs UltrahumanDex · MED
  5. OverDex archive-fusion demonstration — point OverDex at one real, nested folder of mixed multi-vendor exports; show route→run→fuse recovers a coherent multi-signal picture with per-file provenance. The capstone that proves the architecture. real data needs OverDex + migrations · MED
  6. Consumer spirometry repeatability — SpiroDex single-node methods paper: implement ATS/ERS acceptability + repeatability criteria locally, report test-retest (the pulmonary analogue of nights-icc). real data needs SpiroDex · MED–HIGH
  1. Backlog populated June 2026 from PAPERS-ROADMAP-2026-06-24-BRIEF.md (above). The earlier batched re-bundle pass (Integrator sample fix + OxyDex hardening) is DONE — both gates green; see the FIXED findings below. The provenance finding (buildHash covers only the bundle template, not the asset manifest) could still seed a short reproducibility methods note ("what does buildHash actually certify?").

Open findings flagged during analysis (not yet written up)

T © 2026 Michal Planicka ·Tepna v1.0.0 ·Apache-2.0 ·◈ Asheville, NC ·not a medical device