To the Editors:—Of the many wonderful scenes in “Out of Africa,” there is one that applies to Fletcher and colleagues’ article describing an intern’s day on call. During a deluge of heavy rain, Karen Blixen and her Maasai helper are trying to keep the coffee plants from being swept away by forming mud dams to hold the water back. He sees that it is hopeless, and tries to convince her that “…this water wants to live in Mombassa. Let it live in Mombassa.” So it is with medical internships. Not the hopeless part, just the fact that they are difficult to contain or change course. Twenty years ago, we conducted a random work sampling study in an attempt to more systematically document how housestaff spent their time on an inpatient rotation.1,2 We showed that on-call interns spent 11.5 % of their time with patients, roughly 44 % of their time on administrative work, 8 % of their time walking from one place to another, and 6 % of their time on learning activities. Twenty years later, after an extensive effort to change how housestaff are educated, Fletcher and colleagues found that on-call interns spend about 12 % of their time with patients, 40 % on administrative work (all of which is now on a computer), 5 % of their time walking around, and 2 % of their time on learning activities.3 Interestingly, interns 20 years ago slept about 4 h during a 24-h call cycle. In the current study, only half of the interns slept at all, and the average amount of sleep was only about 1.5 h. The majority of residency reform regulations have been directed at enhancing education and improving rest. They don’t seem to be working, assuming that we think they were broken in the first place. An internship is what it is, and it’s a beautiful thing. For me, now about 30 years later, my internship was without a doubt the most important year in my career. I wouldn’t change it one bit, and I wouldn’t trade it for any other year of my career. On a methodological point, Fletcher and colleagues deployed paid, trained observers to shadow interns. We used random work sampling, which relies on self-report. Both seem reliable; random work sampling is cheaper.