The scars
Nobody publishes their failure ledger. We do — because the fixes are the actual product of the first two weeks. Every incident below is verbatim from the burn book; every failure class gets a written postmortem before we move on.
Incidents—logged since day 1
Postmortems—failure classes written up
Crashes—things that died loudly
Money wasted—"waste" — dead burn caught
Data losses—renders we can't get back
Worst day——
The burn book
every row of incidents.json, newest first · host names + stack traces stripped, nothing else touched
The postmortem library
one write-up per failure class (cips.json) — an incident isn't closed until its class has one
Reading order for the morbidly curious: a "repeat" kind means the same class bit twice — those postmortems are the load-bearing ones. The rule of the forge: fail loud, write it down, fix the class, not the instance.