Things break. What you are owed when they do is speed, honesty, and a written account afterwards. This is the process, in the order it happens.
Severity levels
- SEV1 — runtime down or data integrity at risk. Page goes out 24/7; status page within 10 minutes.
- SEV2 — degraded: elevated errors or latency. Status page within 30 minutes.
- SEV3 — narrow impact, workaround exists. Noted on the status page, fixed in normal hours.
During an incident
status.myndlabs.ai is the source of truth, updated at least every 30 minutes during SEV1/SEV2 even when the update is 'still investigating'. We do not mark incidents resolved until we have watched the fix hold, and we do not quietly edit history.
After
Every SEV1 and SEV2 gets a public postmortem within 5 business days: timeline, root cause, customer impact in plain numbers, and what we are changing. No 'a subset of users may have experienced' hedging — if 14% of runs failed for 22 minutes, the postmortem says that.
Security incidents follow the same process plus direct notification to affected workspace owners within 72 hours, as the DPA requires — sooner in practice.