Start with an optical sensor reading. See how a deterministic analysis can interpret the observation without granting authority. Then inspect why a consequential robot effect is not implied by a successful data feed.
Optical raw channels are shown only when the cloud supplies valid counts.
Cloud receive and source-frame digest: awaiting the API.
Source: same-origin GET-only /live/api, based on the IoT.ng cloud. This snapshot is not independent sensor calibration, hardware attestation or a safety monitor.
Inspect live BH1750 Lux and TCS raw RGB/C channels, cloud receipt time, per-frame hash, trend and stale/offline status. A recent accepted cloud reading is still not physical ground truth.
Follow the same observed context into deterministic, explainable light-level interpretations. No AI model inference or learned-model quality claim is made in this read-only demonstration.
Inspect an explicit non-execution boundary. A real sensor frame and a useful analysis are not evidence of a current EG mandate, robot PEP invocation, or realized physical effect.
Proved in the deployed research demo: cloud-accepted optical observations can be viewed through three independent, publicly accessible read-only interfaces with evidence freshness and source-frame hashes.
Not proved by this live stream: calibrated Lux accuracy; independently attested sensor identity or clock; native AI inference; EG current-effect authorization; native Robot PEP commitment; E2E physical-effect reconciliation; or production robotics safety.
Earlier M3 research runs and isolated policy/effect-gate benches are separate evidence scopes. They must not be silently folded into this telemetry run.