First, let me say that none of this is likely to be news to Mitch, but my take is a bit different.

Some bonafides: As an editor and tech writer, I've used Aldus/Adobe products every working day for 15(?) years. I've been sucked into usability work on the product that we make at my current company, and I deal with usability issues every day wrt the customer product information I write.

First, on audience-dependant usability: You must consider your audience when creating an interface (or CPI, or whatever); different audiences have different needs. It may, in principle, be possible to design an interface that works equally well for timid new users and old-hand power users. Doing that while maintaining a consistent interface both within a given release and over several releases is extraordinarily difficult. It's also why Mitch has a job (again). And it's probably directly related to why he was out of work for such a short time.

Some of the specific differences involve how much guidance the user wants or needs (more for new users, nearly none at all for old hand), how many and what sorts of shortcuts (the reverse), and how much error-checking ("Yes, of course I want to do that; that's what I just said.").

A classic example of this is Clippy, the annoying help interface in MS Office. There's actually a market for this sort of help interface, but after a few weeks (or seconds, for some of us), it is far more annoying than helpful. If a new user is to get any benefit from this, it has to be the default, but experienced users would rather never see it. The specific problem here was that MS made it tedious and non-intuitive to kill Clippy dead, dead, dead, not that it was included in the first place. That said, there is a part of the market that would hate it regardless how easy it was to kill.

Second, on Adobe software: The above relates to Adobe products too. There are tools to replace each of them, but every one that have seen has a worse interface for most technical writers and illustrators. The ease of replacement is especially noticeable when you look at packages that don't use proprietary file formats. (For example, a .jpg produced by Photoshop can be read and edited just fine by many other packages.) That said, I wouldn't recommend most Adobe software for new users; to the extent they are optimized for users, it's experienced users they aim at.

Please note that I don't intend the above as ringing endorsements of Adobe's interfaces; there are many places where they could be made better (FrameMaker!). It's just that the same can be said of every one of their competitors. Give me something with all of the capability of Photoshop, Illustrator, and FrameMaker, at the same price, and with a better interface, and I'll jump ship immediately.

Now, in order to have a better interface, you will either have to use something that's recognizably related to the existing interface, or the cost of the changeover (mostly in user time required to adjust to the new interface) will have to be repaid for very quickly.

I see this as directly related to the idea (prominently espoused by Jakob Nielsen for example) that purple is the right color to use for visited links on a website. Objectively, it's not the color I would choose in the absence of history (if only for color-blindness reasons), but most people understand it, and the advantages of shifting to a new color are insufficient to offset the disadvantages in most cases.

As an aside, I've seen brain-development tracts that say that this is the way we tend to work internally, too. Young children go through skill plateaus, then advance remarkably quickly. In most cases, this is the result of discarding a less-efficient paradigm for a more-efficient paradigm, but this doesn't happen until the more-efficient paradigm reaches a critical level of advantage.

Also related to this is that experienced users will never willingly upgrade to a new release or package when close to a deadline. There is always a much larger cost to the upgrade than you expect, and it's seldom something so unimportant as money.

Third, on James Lileks's problems: When Adobe changed GoLive's interface, I suspect it was trying to answer a particular customer issue. While that is laudable, the method was ham-handed. The change was to the effect of a keyboard shortcut, keyboard shortcuts are only useful to experienced users, and experienced users, as Mr. Lileks noted, are overwhelmingly likely to want a new page, not a new site, etc. Had a similar change been made to a "Site-creation Wizard" or the like, it would have much more defensible.

As to Elements, it is, and is intended to be, a crippled version of Photoshop. It wouldn't be more expensive for Adobe to sell full Photoshop than to sell Elements, except that it would kill sales of the much more profitable package. If you buy cripple-ware, it should come as no surprise that its crippled. Before you take that as too much of a snark, I bought Elements myself, knowing that it's crippled, because the price/performance tradeoff was worth it to me. It's still an excellent bitmap editor.

Sorry about the length.