Author here, happy to follow up. We wrote the Meerkat manifesto after our experience Hacker Retreat batch 01. The meerkat method should be more effective than the (few) alternatives we tried. If you hear from people having success doing something similar systematically (pair up; mentor picks a project, gives it to learner, only simplifies if learner is really stuck)... let me know. General lore says that people who pair up more improve faster. The twist here is to pair up with a person who is far more advanced and is invested in getting you up to speed.
My first thought on reading this is that it's counter to what's recommended with Deliberate Practice. This system seems to start with failure and then backs off on the difficulty until the learner succeeds. Deliberate Practice starts within the learner's capability, and pushes outward, trying to stay right at the edge of the learner's capability. I don't see how starting out with failure is going to be very encouraging to the learner.
It's failure in a controlled manner. They know the task is a scorpion, ie, demanding enough for the mentor to not be bored. It could even come out of the Mentor's own daily practice.
Did you read about a 10-year-old who was complaining that calculus (integrals) was hard, but did them anyway thanks to Kahn academy? I think it's mentioned in one of his TED talks. If you don't know that something is too difficult, you may do it anyway.
My experience with say universities is that you get stuck in exercises for far too long. At universities you rarely get to the point of 'project' (unknown solution, creative work needed). At that point they call it research.
I think the weak spot of the Meerkat method is to find a way to keep mentors motivated. For data science retreat, mentors are paid, but they do it because they think it's the right thing to do too. In the wild, one would have to think about why the mentor will want to sit with the learner. One possible way is to let them own the results of the learner's work, and let them work on projects that the mentor had to do anyway. For example, chunks of a freelance job that the mentor had to deliver and could partially unload to the learners. For this, learners must be competent enough to ship production code; which puts even more pressure on both sides, but I think it's the right kind of pressure :)
This is how it worked in middle-age and Renaissance guilds. The master outsourced say 'painting of hands'. The learner did nothing but hands, and the master owned the final work.
Practicing (i.e., getting better at a skill by using it) is rather different from learning/finding out information about that skill. In practicing, you know what you want to achieve; in learning a complex framework finding out what you should achieve is 90% of the problem, and afterwards doing it well is the easy part.
Compare with deliberate practice of, say, a musical instrument - you wouldn't start with the way that the learner initially grasps the instrument and practice that approach (that'd reinforce bad habits, and create problems in future) - instead, they'd go through it with a mentor who'd indicate the failures and show what is the appropriate good form/posture to use instead; and then you practice that. Not try to find out that good form by exploring from your current bad one, but practice the appropriate thing from the very first day - and you likely couldn't tell the difference between the appopriate thing and a bad habit, if it wasn't shown to you. Which is similar to what the article advocates.
More that this method finds the edge of deliberate-practice more effectively than someone trying to figure it out on their own. The iterative process causes a rapid calibration between teacher and learner, until the just-hard-enough level is found. Rinse and repeat as the learner progresses. Variations of this approach have been used in sports, music, flying, etc.
i'm actually doing the mentor role atm for a couple of people, but in one or two of the situations I can't really send them work home to do on their own.
So i'm taking the active role in the pair programming, and showing rather than having them doing. At least they will have the reference afterward.
gitbook looks like a pretty nice way to collect all that knowledge though.
