For a long time, an error in Freebo was a note to myself. Something broke, I noticed eventually, I fixed it, no one else ever knew. That is the quiet privilege of not having customers.
I have a customer now. A live tour operator, taking bookings. So an unhandled error is not a note to myself anymore. It is a customer who cannot check out, an operator who loses the sale, and a very reasonable conclusion on their end that the tool does not work.
Launch changes who pays for a bug
This is the shift nobody warns you about. Before a real user, a silent failure costs you a little cleanup later. After, it costs someone else money in the moment, and it costs you their trust, which is the only thing a new platform actually runs on.
So this week I did the unglamorous launch work. I wired production into Sentry so errors page me instead of hiding, and I tightened the Stripe paths a booking depends on. Not to impress anyone. Because a bug has a victim now, and I would rather hear about it from a monitor than from an operator asking why their customer walked.
Observability is a trust decision
I used to think error monitoring was a “real engineer” checkbox, the kind of thing you add when you feel like being professional. It is not. It is the moment you admit failures are inevitable and decide who finds out first: you, quietly, at 2am, or your customer, publicly, at checkout.
Launch is not when you stop building. It is when the cost of a silent failure changes owners. Everything I add now, I add because there is finally someone on the other end who would feel it break.