Nobody tells you about the ledger on day one. Or the rounding rules. Or why moving £0.01 to the wrong account is a compliance event. I learned all of it the hard way, and it took longer than it should have because nobody warned me what kind of software I was actually building.
No one told me the button actually pushed to live. I found out when my manager went pale. What followed taught me more about engineering culture, incident response, and deployment design than anything I learned in the next six months.
You click 'Place Order' and a spinner appears. Behind that spinner, a surprisingly complex distributed system is racing to confirm something that might not even be possible, and the outcome depends on a chain of systems that don't share a database and can each fail independently.
Four interviews, four rejections, all with variations of the same vague feedback: didn't go deep enough. I finally understood what that meant on attempt five, and it had nothing to do with knowing more architecture patterns.
The query ran in 3ms on my laptop. In production, against three years of real customer data, it took 47 seconds. Understanding why taught me more about databases than any course I had ever taken.
I thought I understood the domain because I had used hospitals as a patient. I was wrong about almost everything. Eight months of working on clinical software taught me that the invisible assumptions in healthcare IT are denser than in any other domain I've worked in.