Vercel has a control plane API that for circular dependency reasons cannot itself be hosted on Vercel. It's a k8s cluster fronted by Vercel's primary ingress service using special hard coded rules. Long-story short, that API routing might be the most core of our core infra. Well, our 1st time intern @s4chinraja went ahead and said: Bundling API routing with the hyper scale ingress service that has to roll out very carefully isn't the right architecture. In week ~2 he had a RFC reviewed in our weekly RWC (Response with Comments) design review, then he built the thing, and last week, well before the end of his internship, he fully rolled out the new dedicated API routing service globally to 100% of traffic. In my whole career I've never seen anything like it. Open heart surgery on the deepest infrastructure, that other folks wouldn't event touch, with full production deployment is just absolutely stunning.