Under the hood
How it works
What the pipeline would do end to end, and — stage by stage — which parts actually run today.
- 01
Acquisition
Planned, not builtA supported sensor would sample a biological signal at a stated rate, with stated units and known error characteristics. No such sensor is connected to anything on this site or in the current apps, and no sensor adapter has been written.
- 02
Feature processing
Planned, not builtSelected features would be computed from the raw signal over defined windows — each with a written definition, unit and window length. Which features, and how they are combined, is exactly the work that has not been specified yet.
- 03
Normalization
Planned, not builtFeatures would be mapped onto the bounded 0–1 scale against a stated reference, so that “low” and “high” mean something specific rather than something relative to whatever happened a moment ago.
- 04
Quality handling
Planned, not builtSignal quality is a separate question from signal magnitude. If contact is lost or the signal is unusable, the correct output is unavailable — not zero, and never a simulated value quietly substituted to keep the picture moving. A gap must be visible as a gap.
- 05
The control value
Built todayA single bounded value in 0–1, displayed as 0–100. This exists and works. Its source is tracked and labelled as one of
simulatedManual or simulatedAuto— there is no third option pretending to be a sensor. - 06
Rendering
Built todayThe value drives Reality Modulation’s colour pipeline: a luminance S-curve, a gradient map onto the selected look’s palette, a blend against an enriched version of the real scene, then saturation, vibrance, bloom and — for the looks that use them — lit edges. 8 RM looks are available here.
Effect strength multiplies the value to cap the result, so the transformation can never exceed what you allowed. Below a threshold of 0.001 the pipeline returns the source image untouched.
- 07
Glasses output
Planned, not builtDriving RM eyewear from this value is a product intention, not a shipped capability. Nothing here has been verified against physical hardware.
On the numbers. Timing uses a monotonic clock, so the simulation cannot be advanced by changing the system time and does not depend on frame rate. Auto interpolates with a smoothstep, giving zero slope at both ends of each segment: the value never snaps, and there is no flashing or rhythmic pulsing anywhere in the design.
On honesty in the code. The same constants drive the web preview and both native apps, from one shared contract file, and the same numerical fixtures are asserted on every platform. That is deliberate — a control value that meant different things in different places would make every claim on this page unverifiable.
The concept
A supported sensor measures a biological signal. Selected features of that signal become one bounded, normalized control value. That value drives how strongly a scene is transformed.
What is built today
The colour pipeline is real and running. The input is not: every value here is either a slider you move or a synthetic signal generated on your device. No sensor is connected and nothing measures you.