Cal Newport on "Write Every Day". If it's not your main job, you're going to end up having no write days, and if you're committed to a hard schedule a missed day is going to translate into "welp, couldn't make the cut then, better quit for good".
Yes, The Mind Illuminated is basically the same ten-step model as the one in that article, but expanded to book length and with lots of extra practice advice and theory of mental models.
The Mind Illuminated by John Yates is my new favorite meditation instruction book. Has lots of modern neuroscience grounding, completely secular, and presents a very detailed step-by-step instruction on going from not having a daily meditation habit going to attaining very deep concentration states.
One problem is that the community has few people actually engaged enough with cutting edge AI / machine learning / whatever-the-respectable-people-call-it-this-decade research to have opinions that are grounded in where the actual research is right now. So a lot of the discussion is going to consist of people either staying quiet or giving uninformed opinions to keep the conversation going. And what incentive structures there are here mostly work for a social club, so there aren't really that many checks and balances that keep things from drifting further away from being grounded in actual reality instead of the local social reality.
Ilya actually is working with cutting edge machine learning, so I pay attention to his expressions of frustration and appreciate that he persists in hanging out here.
This sounds like someone who's salient feature is math anxiety from high school asking how to be a research director at CERN. It's not just that the salient feature seems at odds with the task, it's that the task isn't exactly something you just walk into, while you sound like you're talking about helping someone overcome a social phobia by taking a part-time job at supermarket checkout. Is your friend someone who wins International Math Olympiads?
Maybe someday someone clever will figure out how to disseminate that knowledge, but it simply isn't there yet.
Based on Razib Khan's blog posts, many cutting edge researchers seem to be pretty active on Twitter where they can talk about their own stuff and keep up on what their colleagues are up to. Grad students on social media will probably respond to someone asking about their subfield if it looks like they know their basics and may be up to something interesting.
The tiny bandwidth is of course a problem. "Professor Z has probably proven math lemma A" fits in a tweet, instruction on lab work rituals not so much.
Clever people who don't want to pay for plane tickets and tuition might be pretty resourceful though, once they figure out they want to talk with each other to learn what they need to know.
And for people on the Vim side, there's VimOutliner for doing workflowy-like outlines, also with a time-tracking component.