Your dilemma is yours.

No sign-in is needed. Questions and conversation histories are held in browser memory. Decode attempts and hint exposure use tab-local session storage; Compose drafts and ten revisions use device-local storage until deleted. Content is sent to the configured AI provider for inference. They are not persisted in the application database, analytics or URLs. Clear the conversation or close the tab to remove local context. The operator must disclose its selected model provider and contractual retention terms before public launch. Anonymous browser measurement is off by default; a previously granted session preference remains valid until the tab closes. Server operational metrics record event names, controlled topic categories, language, verse IDs, timing, token counts and estimated cost. No raw questions, attempts, IP addresses or histories are stored in analytics. A transient HMAC of the client address may be used for rate limits; raw addresses are not retained by this app. Hosting providers may separately process access logs; operators must apply the deployment log exclusions. Analytics defaults to 30-day retention and a minimum group size of five. The maintenance job must run daily. Feedback is optional, redacted for common email/phone patterns, held for moderation and never published automatically. Do not include sensitive personal details in feedback. Administrative analytics access is audited.

Your writing and learning attempts

Use Delete local draft in Compose or clear browser site storage to remove drafts and revisions. Decode attempts are separated by Kural number and removed when the tab session ends or site storage is cleared. Craft feedback sends the current composition to our configured Vertex AI provider only when requested. The application does not store draft or answer text in analytics. Device-local drafts are available to anyone using the same browser profile; there is no account-based cloud draft sharing. Activity counts do not measure proficiency or learning outcomes.