AVAM is governed by a clear promise: intelligence must remain accountable to the human life it supports. Each principle below is connected to a verified operational control—not a stated aspiration.
These are operating constraints, not marketing claims. Every feature and integration must earn its place against them.
AI advises, organises, and acts within delegated boundaries. Material choices remain understandable, interruptible, and controlled by people.
LiveEmail drafts require Mike to review and send before transmission. High-impact actions are gated in code. Automated rollback is threshold-governed and human-notified.
Information is used for the specific purpose it was entrusted for. New uses require a conscious decision, not silent expansion.
LiveLarry's database uses a context column (personal / official) to separate data domains. PlannedFormal data retention policy governs deletion schedules.
Privacy is the starting architecture. Access is limited, sensitive contexts are separated, and exposure is minimised.
LiveSingle-user system; no public accounts. TOTP multi-factor authentication on admin access. No analytics, tracking cookies, or monitoring scripts on this website. PII masked before vector embedding.
Personal data exists to create value for the person it describes. It is not a commodity and is never the hidden price of convenience.
LiveHealth and biometric data stays within Larry's private database. No personal data is sold or shared. AI observability logs are on a self-managed private server under Mike's direct control. PlannedAutomated retention purge schedule.
Important actions should be explainable. Sources, assumptions, and automation boundaries must be visible enough to support informed review.
LiveThis website publishes a capability register classifying every feature as Live, In Development, Planned, or Illustrative—based on a September 2026 source-code audit.
AVAM uses only the level of access and automation the task genuinely requires. Capability does not automatically create permission.
LiveEmail is read and drafted only—never sent autonomously. Calendar is read-only. Code execution is sandboxed with 30+ forbidden patterns.
Protection must be resilient in daily use. Reliable operations, careful monitoring, and recovery planning reinforce trust over time.
LiveSelf-heal monitor runs every 5 minutes with 13+ checks and auto-rollback. TOTP MFA, session expiry, and CSRF protection active. Weekly automated code security review. PlannedAutomated database backup.
Success is measured by greater clarity, wellbeing, contribution, and connection—not simply by more automation.
LiveHealth monitoring includes a medical disclaimer and does not replace professional care. AVAM does not optimise for engagement or data collection—it optimises for Mike's clarity and purpose.
If a capability cannot be used with dignity, transparency, and genuine benefit to the person, it does not belong in AVAM.