Reliability priorities
The profile spine is the highest-priority state because everything else reads from it.
Tracked jobs, wins, reminders, and mode changes should survive transient failures cleanly.
If AI-backed surfaces struggle, Dilly should fail legibly instead of acting like the user did something wrong.
When reliability shifts, the status surface should reduce confusion before support tickets pile up.
What matters most in a connected system
Traditional uptime metrics miss the real Dilly experience. A jobs page can render while still being wrong for the user if background context is stale. A resume generation surface can be technically “up” but operationally degraded if grounding is thin or retries are spiking. Reliability has to be measured around useful continuity, not just request success.
Practical principles
- prefer preserving user state over preserving cosmetic polish
- replay background work when recoverable instead of silently dropping it
- make stale data visible where it changes user decisions
- treat reminders and calendar sync as trust surfaces, not minor side features
- log mode transitions and profile mutations as system-critical events
Why this helps the company feel real
Companies feel legitimate when they communicate operations openly. A status surface tells users, partners, and future buyers that Dilly expects to matter enough for service health to deserve a public room.