What should I do if I make a serious mistake at work?

Report it immediately, focus on fixing it before explaining it, and afterwards address the systemic cause rather than just apologising. How you handle a mistake is scrutinised far more than the mistake itself — and handled well, it can genuinely increase trust rather than reduce it. **In the moment, in order:** 1. **Tell someone now.** The instinct to quietly fix it first is the single worst move available. Delayed disclosure turns a technical problem into a trust problem, and trust problems are the ones that end careers. Two hours of downtime that you flagged in minute one is a different event from two hours you tried to hide. 2. **Stop the bleeding before diagnosing.** Roll back, disable the feature, restore service. Understanding *why* comes after the users are fine. 3. **Communicate status clearly and frequently** while fixing. 'Rolled back, service restored, investigating cause.' People tolerate problems far better when they're informed. 4. **Ask for help.** Panic makes you slower, and a second person catches what you'll miss. **Afterwards:** 1. **Write an honest, blameless post-mortem**: what happened, timeline, impact, root cause, and what changes prevent recurrence. Focus on the *system* — why was it possible to deploy this without a check catching it? — rather than on your own inadequacy. Self-flagellation is uncomfortable for everyone and produces no improvement. 2. **Propose the fix that stops it happening again.** A test, a staging check, a deploy gate, a review requirement. Turning your mistake into a permanent improvement is what converts an incident into a contribution. 3. **Apologise once, briefly, and then stop.** Repeated apologising makes colleagues manage your feelings on top of the incident. **On the terror**, which is the real subject of your question: essentially every experienced engineer has broken production. It's close to a rite of passage, and the ones who haven't have usually just not been given enough responsibility yet. The stories get told at length precisely because they're universal. **What healthy organisations actually do**: treat incidents as system failures, not individual ones. If one person's single mistake can take down production, the deficiency is in the safeguards. Any team that responds to this by punishing you rather than fixing the pipeline has a culture problem — and that's genuinely diagnostic information about whether you want to stay. The realistic outcome, if you handle it as above: within a month it's a story, and your colleagues remember that you flagged it immediately and fixed the underlying gap. That reputation is worth more than never having broken anything.