I completed CS373, "Programming A Robotic Car", at Udacity in April, and can answer these questions.

There is generally one lecture and one homework exercise posted each week. The deadline for the homework is the following week. You can resubmit homework any time before the deadline, and see how you did any time after the deadline. There are also "quizzes" within the lectures: unassessed exercises for which you get the answer as soon as you want to look at it. There is a final exam at the end (isomorphic to "homework", just called by a different name). You get a digital certificate at the end if you do well enough on the homework and exam. Behold!

Once material is posted, you can see it until the end of the course (maybe afterwards indefinitely, I can still see them when logged in to Udacity). In fact, I can see all the materials for courses I haven't done as well, including answers to exercises and exams, which is surprising. But maybe I'm looking at the exercises and answers from previous runs of the courses and they'll be different the next time round.

For courses being run for the first time (such as the suggested Statistics course), all I can see is the preview video -- it's quite likely that all the lecture and homework materials haven't been completed yet.

So I suggest you register at Udacity and look around to see what you can see.

It's up to you to decide how you want to work with this. If you're not interested in the certificate, you could just register at Udacity, wait until the course is completed, then work through all the material at whatever pace you want. Something you will lose by doing that is the interaction with other people doing it in real time. You can read the Udacity forums after the fact but you wouldn't be participating in them.

Personally, I preferred to work through the course in real time, meeting all the assessment deadlines, but YMMV.

I recommend Old Papers.

It's curious that in all three volumes of that work (searched on Amazon), there's not a single paper by E.T. Jaynes, and just a single mention of him anywhere.

In old papers people know what they are talking about, and how it relates to everything else from their time, and how it relates to what their innovation is, and they explain it all.

Presumably implying that people don't do that now? Is that a general trend, and in other fields than statistics? To take a nonrandom example, Judea Pearl seems to measure up well by those standards.