# Noetryx technology brief Updated 11 October 2026. ## The pitch Noetryx turns measured cognitive practice into a traceable learning loop. Its software agents connect task records, an inspectable personal model, a next-session plan and fixed-protocol review. ## Audience The primary audience is individual adults worldwide: everyday users, students, professionals, athletes and gamers. Esports is one use case. A Free/Personal membership page displays proposed $9.99 monthly and $79 annual pricing; the implemented personal program remains free early access with no checkout or automatic renewal. Personal subscriptions are a proposed revenue model; coach and organization plans are future extensions. The current English release is publicly accessible, with sign-in required for saved personal records; billing, group access and paid adoption have not been established. ## Working software Four browser tasks, changing-target Focus Grid, per-trial records, shared browser/server scoring, private persistence, immediate rule-based guidance, an on-demand four-agent service, one-click preparation of the next session, model evaluation and exportable decision records. The personal program also saves a goal, context, time budget, practice days and time zone, prepares the next session, and calculates seven-day reviews from comparable fixed measurements. At least three fixed sessions on two dates in each window are required; this rule is not a clinical validation threshold. ## Agent architecture 1. Signal: recalculates measured trials and selects comparable current-protocol records. 2. Model: fits a standard ridge regression after at least 20 comparable sessions over five dates. Rolling chronological validation compares against a prior-three mean; no forecast is shown unless model MAE is lower. 3. Practice: proposes a baseline, practice or fixed recheck based on explicit policy. For choice/control, establish three fixed sessions, then recommend at most two saved practice sessions before another fixed check; accuracy below 75% returns to a careful fixed session. This is an experimental product policy, not a validated dose or efficacy claim. 4. Review: compares the first and latest three fixed sessions only when the date windows are separate. Speed and accuracy changes remain separate. The agents are bounded software components, not autonomous chatbots. Only the Model agent learns numerical weights. No external AI inference API is required. The forecast does not control the practice prescription. ## Original implementation and dependencies The task code, scoring implementation, regression implementation, agent orchestration, planning policy, API, storage and interface are stored in the Noetryx project source repository. Ridge regression and behavioral-task concepts are established methods. Frameworks and hosting retain third-party dependencies and licenses. This does not establish patent protection, scientific exclusivity or legal clearance of a brand. Private user task records are separate from the nonexclusive public physiological research dataset. The public experiment is not the personal agent's training dataset. ## Why a user may value it Each test gives an error-specific practice priority. Saved history produces a next step with inspectable reasons. Fixed assessments help users distinguish a changing challenge from a changing result. The system shows failed model comparisons and speed–accuracy trade-offs. These are implemented workflow benefits. Improved task performance, transfer, retention and clinical effects have not been established in prospective Noetryx studies. ## Capability status - Performance scenarios: four behavioral tasks implemented. - Telemetry: behavioral records and local EEG recording/secure-stream processing implemented. A physical source must be configured separately. - AI engine: experimental personal regression and agent orchestration implemented. - EEG signal engine: original Welch processing, quality checks, band power and calibrated visual feedback implemented; physical-device verification pending. - Adaptive metrics: task timing, accuracy and errors, with separate EEG spectral measurements; no brain-state or IQ score. ## Defensibility thesis and honest pitch boundaries Our chosen design distinction is a traceable chain from measurement to quality checks, learned-model evaluation, a bounded plan and fixed-protocol review. This is not verified worldwide uniqueness. Adaptive training and baseline tracking already exist, including in BrainHQ and NeuroTracker: - https://www.brainhq.com/why-brainhq/brainhq-faq/ - https://www.neurotrackerx.com/post/guide-neurotracker-scores Establish advantage through reliable measurements, permissioned longitudinal data, prospective outcomes, user retention and paid demand. No achieved investment valuation is asserted. A $100M valuation is an ambition, not an evidenced current fact. ## Signal Engine update A browser EEG processing engine now supports local CSV analysis, Welch spectral measurements, per-channel quality flags, a secure live acquisition receiver, and calibrated alpha-power visual feedback. It requires real user-supplied recordings or a physical acquisition bridge. No hardware test or clinical benefit is claimed. The personal task model remains separate. See [the EEG protocol](/evidence/EEG-ENGINE.md) and [bridge setup](/evidence/EEG-BRIDGE.md).