Trust center
Progress has to be earned by evidence.
CloudKata separates motivating game state from readiness, keeps source history append-only, and labels uncertainty instead of presenting a score as certainty.
Effective July 13, 2026 · Private beta
Readiness boundaries
- Game rewards and streaks never feed the readiness projector.
- Reading a guide or writing a lab reflection does not increase correctness.
- Scored observations retain content version and server-derived taxonomy attribution so projections can be rebuilt and audited.
- Low evidence is shown as insufficient evidence, not confident mastery.
Account and data controls
Learner rows are protected by database row-level security. Completion actions are idempotent and transactional, and export/deletion cover the registered learner-data inventory. Privileged maintenance and content actions are separately authenticated.
Tutor guardrails
The tutor is grounded in server-loaded study sources, reconciles citations to retrieved material, marks missing coverage as uncertain, refuses prohibited exam-dump requests, and cannot change a study plan without explicit learner confirmation. Provider calls have per-request, daily, monthly, and global cost ceilings.
What is still limited
This is a small, allowlisted private beta. Platform logs are the only live operational signal, no managed database backup is configured, and Kubernetes browser terminals remain preview-only. Managed observability, custom SMTP, backups, restore rehearsal, and human accessibility review are gates for external beta.