An evidence chain, not a black box.
The current development system is built around deterministic telemetry processing, explicit data quality and a narrative layer that is not allowed to become the source of measured facts.
Ingest
Normalize the recording
Condition
Clean the signal
Track model
Resolve laps and corners
Analysis
Decompose the time loss
Narrative
Explain what's grounded
Multiple recording ecosystems
Development support includes VBOX-style telemetry, RaceCapture, TrackAddict/RaceRender, Garmin FIT and GoPro telemetry paths, with device-specific handling where timing and sensor semantics differ.
Signal quality before interpretation
Despiking, orientation, resampling, synchronization and sensor-quality checks happen before higher-level claims are made.
Laps, corners and comparable distance
Sessions are converted from raw time-series recordings into track-relative structures that can support meaningful comparison.
Time-loss decomposition
The core engineering target is not simply identifying a slower corner; it is reconciling the measured lap difference into useful, testable causes.
Grounded explanations
Measured findings cross into the narrative layer with references. Unsupported numeric claims are rejected rather than polished into plausible-sounding advice.
Unavailability is a product state, not an error to hide.
If the evidence cannot support a claim, SectorZed should say so. Missing or unreliable data is not zero, and uncertainty should not be styled away.