Protocol v3.2

How we
protect flow.

We do not sell focus apps or blocklists. We build lightweight models that learn your team's cognitive rhythm, then quietly deflect interruptions before they fracture your attention. The result is a measurable reduction in context-switching cost without adding friction to your workflow.

Last reviewed JUNE 2026

Core protocol values

Telemetry by consent

Our systems operate with minimal data collection. We do not monitor keystrokes, capture screen content, or build productivity scores for managerial review. All behavioral signals are anonymized and processed locally by default, with cloud options available only under explicit contractual terms.

Signal over noise

We distinguish between urgent interruptions and deferrable noise using a multi-factor model that weighs sender history, message semantics, and your real-time cognitive load. False positives are logged and reviewed weekly by our research team to improve model calibration.

User retains control
Every intervention is transparent and reversible. You can inspect why a notification was delayed, override any decision in the moment, and adjust your protection thresholds without filing a ticket. The system serves your focus, not an employer's surveillance agenda.

Human in
the loop.

Our automated models handle the high-volume, low-stakes routing decisions that consume mental bandwidth. When the system encounters an ambiguous signal or a novel interruption pattern, it escalates to a human reviewer rather than guessing. This hybrid approach keeps error rates low and maintains trust in the system's judgment.

Human reviewers are trained cognitive researchers who meet weekly with the engineering team to refine edge-case handling. Their annotations feed directly back into model retraining, creating a continuous improvement loop that pure automation cannot replicate.

< 2% Escalation rate
48h Review turnaround
Cognitive researcher reviewing model outputs on a screen

Validation metrics

Control criterion Minimum threshold Audit method
False positive rate < 5% Weekly manual review of delayed notifications
User override rate < 10% Monthly survey of active users
Focus session completion > 75% Self-reported flow-state diary correlation
System latency < 200ms Continuous automated endpoint monitoring
* All thresholds are reviewed quarterly against anonymized aggregate data. Individual client results may vary based on integration depth and team communication patterns.

See if your workflow qualifies

We begin every engagement with a technical feasibility review. Tell us about your stack, your constraints, and the last time an interruption cost you a clear thought.
d

Syfraurtiosrobrilda research unit — Vancouver

  • Peer-reviewed methodology
  • Privacy-first by design
  • Continuous model validation