The Foundation of Everything
The Sensor is the first layer of the 5-Layer Architecture and the foundation of everything that follows. Its principle is simple but radical: every signal that can inform a decision must be captured before it becomes dark data. No sampling. No batch aggregation. No selective collection based on what someone decided was important last quarter. Total information capture.
Protocol-Level Capture
The Sensor operates at the protocol level - below the application layer where most analytics tools collect data. This means it captures signals that application-level tools miss: network timing, micro-behavioral patterns, cross-signal relationships, and sequence context. A click is not just a click - the Sensor captures the hesitation before the click, the scroll pattern that preceded it, the content that was visible at the moment of interaction, and the signals from other users viewing the same content simultaneously.
Multimodal Input
The Sensor processes multimodal input across video (object detection, scene understanding, visual attention patterns), audio (tonality, stress levels, enthusiasm indicators), text (language, keywords, intent markers), and visual semantics (what users see and how they respond to visual elements). This multimodal capture is what enables the Translator to produce rich intent classification rather than simplistic click-based attribution.
2.4M Signals per Second
Processing 2.4 million signals per second continuously is an engineering challenge. The Sensor uses a streaming pipeline architecture that processes signals as they arrive rather than batching them. This stream-first approach ensures that the Translator and Logic Engine receive signals in real time - no batch lag, no dark data window. The architecture is designed for this: Kafka-based streaming, container-based elastic scaling, and priority routing through Nami.
Context Preservation
Raw signals are meaningless without context. The Sensor preserves session awareness (which signals belong to the same user session), cross-signal relationships (which signals are responses to other signals), sequence metadata (in what order did signals occur), and temporal context (exact timestamps for downstream Temporal Resonance analysis). This context preservation is what makes the Logic Engine's contextual weighting possible - without it, the same signal would always receive the same weight regardless of circumstances.
No Dark Data
The Sensor's design philosophy is that dark data is a design failure. If a signal was generated and could inform a decision, it should be captured. The cost of processing signals that turn out to be irrelevant is lower than the cost of missing signals that would have informed a critical decision. The 5-Layer Architecture downstream handles relevance - the Sensor's job is comprehensive capture.