If you are 'learning' stuff open-loop (with no feedback from users or peers), then you are deluding yourself about how much you've actually learned. You're one important but terribly insufficient step above book learning.

No battle plan survives contact with the enemy. There are tons of strategies I know to solve problems that I never use because it confuses other people (and some of them confuse me too when I look at them six months later). That's experience. That's wisdom. You won't find that in a private github repository, and only those who already know what they're looking for will find it in a book.

> If you are 'learning' stuff open-loop (with no feedback from users or peers), then you are deluding yourself about how much you've actually learned. You're one important but terribly insufficient step above book learning.

Book learning ain't bad; in my experience, most of the developers I know would do well if they picked an actual book every now and then. It's not as good as hands-on experience, but still miles ahead of "learning" by StackOverflow-driven development. Experience is an important piece of the puzzle of wisdom, but it's not the whole thing. You need to also have understanding, and reading (and thinking about what you read) is a good way to acquire insight.

WRT. closing the loop, your users and peers aren't the only way to close a loop (and arguably, they can only provide certain kinds of feedback that are very context-specific). Trying things out for yourself is also a good way to close the loop through feedback from reality itself (e.g. whether something works, or how difficult you find it, are such pieces of feedback). On top of that, peer/customer feedback makes no sense for a lot of useful knowledge - for instance the new ES2018 and CSS features you've mentioned upthread.