The big migration left something behind that outlasted it: the Behat test suite built to keep it honest didn’t get archived with the project — it became the standard across the shop. Every migration after inherited the guardrails. This chapter is about what that taught me, because it reshaped what I think the job is.

The insight is simple and it took me a decade to actually believe: untested work is a rumor. Not bad work — plenty of untested work is excellent. A rumor: a claim with no way to check itself, whose truth depends entirely on who’s telling it and how tired they were. A test converts the rumor into a record. “The migration preserves every relationship” stops being a thing the lead says and becomes a thing anyone can run — which changes it from trust in a person to trust in a system, and systems don’t get tired, quit, or leave with the knowledge.

Spreading the suite across a shop taught the second lesson: process adoption is a product problem. Developers don’t adopt testing because a wiki page said to; they adopt it the first time it catches something that would have embarrassed them, personally, in front of a client. So you don’t evangelize — you instrument the next real project, wait for the save, and let the save do the preaching. Culture follows receipts.

And it recalibrated how I measure my own contribution. Code I’ve written gets replaced — should get replaced; everything modern is future-legacy. But a practice installed in a team keeps compounding after you’re gone. It’s the taper’s logic: the show ends, the tape circulates, and the music keeps teaching people who never saw the band. By this point in the trip I’d started judging engagements by what would still be true five years after I left the room — a metric that, looking back, was quietly aiming me at running my own practice.

Next: Back to school — finishing the degree eleven years in.