This seems likely to be controversial but I want to put forward "sales". Every so often I wonder if I should spend several months in a job like selling cars, where things are presumably really stark, but so far I've generally ended up doing something more kosher and traditionally "geeky" like data science.
However, before I knew a marketable programming language I had a two separate "terrible college jobs" that polished a lot of stuff pretty fast: (1) signature gathering for ballot measures and (2) door to door campaigning for an environmental group.
Signature gathering was way way better than door-to-door, both financially and educationally. Part of that is probably simply because there were hundreds of opportunities per hour at peak periods, but part of that might have been that I was hired by a guy who traveled around doing it full time, and so he had spent longer slower cycles leveling up on training people to train people to gather signatures :-P
Sales is objectively measurable, but it is not easily measurable. At least, for big ticket items like cars, the sample size is small so there is a lot of noise. It is easy to measure whether one person is better than another, but it is harder to measure continuous progress.
I credit an undergrad summer job in door-to-door sales for moving my social skills from "terrible" to "good". For that particular job we literally had a points system that was visible to everyone in the office (and determined incentives like fully-paid vacations abroad), and you'd sell enough on a daily basis that you knew roughly how you were doing (ie 5 sales was a decent day, 10 outstanding, 2 bad, out of perhaps 100 interactions), so it was a near-perfect training ground.
Every so often I wonder if I should spend several months in a job like selling cars
I did just this. It was rather difficult, but my confidence and communication skills did improve quite a bit.
Dual N-back
A Lung improvement machine $11
Sprinting
See The First 20 Hours: How to Learn Anything . . . Fast! by Josh Kaufman, whom I met at a CFAR workshop.
Nice recommendations, thanks.
Going from dual 2-back to dual 3-back was pretty hard! I had to (1) completely ignore audio, master 3-back with visuals only, (2) completely ignore visuals, master 3-back with audio only, then (3) try to combine them.
Math textbooks - you read the chapter, learn how to solve the problems, then attempt them and see if your answer matches the ones in the back of the book. Having a professor give you tests and grade them is optional but often useful.
If you already know parts of the material, you may want to take the tests provided in the textbook first, and then focus on the material associated with the problems you did not solve correctly.
Does it have to be something where the progress is easily quantifiable, or merely very easily observed? Learning to program is something that has an extremely long tail of improvement, but the first part goes very quickly. It may not be easy to quantify your progress - rough metrics like "lines of code to solve this problem" or "the amount of computer time / memory needed by this program" do exist, but they are at best only approximations of skill - but if you look back at code you wrote even a month ago (I'm assuming a not-too-strenuous instructional course, here) you'll be shocked by how quickly you improve. If you already know how to code, the same effect (and much of the same benefit) can be obtained by branching out into a really new language, ideally with a totally different coding paradigm; if you know C++, Java, and Python, try learning a functional language like Haskell of F#. If you've never worked "close to the metal", try picking up C and then C++. If you've only ever written application code, learn some scripting language(s) and write little tools to do things you currently do with a GUI file manager / config file editor / registry editor.
Always keep track of what you've written before (if you're doing this as part of a course, old assignments will suffice) and then go back later and write it again "correctly" using your current knowledge. It's entirely possibly you'll see obvious improvements week over week. Heck, when I'm learning a new language if I write 200 lines of code in it without going back and editing more than strictly necessary, the bottom half will be visibly better than the top.
Juggling
Touch typing
Martial Arts (if you count fluid movement and strength and such as measurable progress, rather than viewing the goal to be "combat effectiveness" or somesuch)
Speedcubing. I don't recommend it, though—I started about a year ago and it sniped a significant amount of my free time in 2014, on the order of 400-500 hours. (I had a similar experience with Go in college.)
If you are a Duolingo user, you can use their metrics (levels completed, total points, length of streak, percent of text readable, etc.).
Ability to consume media in that language. For instance, I can read more Le Monde than I can Die Zeitung.
Debatable how measurable progress in meditation feels, but when I've stuck with it in the past, I've subjectively felt noticeable progress.
For programing there are a few quick feedback programs:
code.org for beginners
codility lessons for advanced stuff
I would say consuming mass amounts of media, learning an instrument, talking with other people, and spending time observing.
I am definitely interested in getting better at both "talking with other people" and "observing"; how would you measure your progress in these two cases?
I used to have severe social anxiety. A lot of factors helped me get over it. But talking to people was definitely up there. I'm not scared of people today, but my social skills are still a bit lacking.
I would setup metrics of success like length of conversation, quality (which is subjective), and if you traded information to further stay in touch. Observing people wouldn't have such metrics. It is a discipline that you simply must practice. The key would be not to get distracted. You may wish to time yourself and set a goal to do it a certain number of times a week. It also helps with situational awareness which helps with being safer in public and being ready for a crisis should one arise.
Most of the less desirable skills have highly measurable progress and tight feedback loops. Picking fights, public intoxication, alienating friends with boorish slurs, etc. Perhaps something worth doing can be known in part because it does not come easily.
A lot of my significant personal improvement happened as a result of highly measurable progress and tight feedback loops. For example: