No thanks.
To take the above points:
1. Down time is useful for, oh, things like co-ops.... It's also part of the research university way of doing things, which 7 wants to abolish.
2. Some places like MIT let you test out of many/most classes; if appropriate, you can take the final and that counts, whatever the result.
3. "Mandatory, off-topic courses" are all part of the ideal of a liberal vs. vocational education. While I feel that's been mostly perverted into modern liberal/radical indoctrination and note that outside of MIT and Caltech the reverse doesn't hold (as of the late '80s the math "requirement" at Harvard was proving you could do algebra), I strongly believe there's value to it and would prefer reform to abolishing the concept if that looks at all practical.
4. From my science, CS, programming and sysadmin background I can't comment on this one.
5. Offline is an increasingly interesting option, but I think we're a long way from it being a substitute for good versions of the real thing (a involved student body, people with a clue doing the teaching, etc.; plus there's the issue that US broadband is a less than ideal platform to attempt this from).
6. Co-op is again vocational in focus, and one wonders if the author is aware that taking all this too far will be eating our seed corn by not producing the next generation of academics.
7. In my and my friends experience, a real world class researcher who's also serious about undergraduate education (if not always entirely competent in doing it) is vastly superior to mid-tier school professor who's research is nothing to write home about and who envies his betters. At the "bottom", the community colleges and their one step about 4 year equivalents where the professors are their to teach with no pretenses about doing serious research can provide some truly excellent education in well established areas.
But with the following caveats:
In the case of those needing post-graduate education (e.g. for those fields where a Ph.D. is the minimal level of education), the single most important thing is that those doing the graduate school admissions to know who's recommending a student.
In new fields like CS where we don't agree on which end is up, the results are likely to be very mediocre. E.g. do you want to go to a Javaschool or one which will teach you the foundations that will allow you to learn the languages and paradigms de jour as needed?
Bottom line: fix what's broken rather breaking everything and starting from scratch. There's going to serious leverage here as the US higher education bubble starts popping and I suggest ruthless use of that.