I was reading BusinessWeek when I stumbled on to a bit of a treasure.

Interesting presentation at SXSW from Michael Lopp, senior
engineering manager at Apple, who tried to assess how Apple “gets”
design when so many other companies try and fail. After describing
Apple’s process of delivering consumers with a succession of presents
(“really good ideas wrapped up in other really good ideas” — in other
words, great software in fabulous hardware in beautiful packaging), he
asked the question many have asked in their time: “How the f*ck do you
do that?” (South by Southwest is at ease with its panelists speaking
earthily.) Then he went into a few details:

Pixel Perfect Mockups
This, Lopp admitted, causes a huge amount of work and takes an enormous
amount of time. But, he added, “it removes all ambiguity.” That might
add time up front, but it removes the need to correct mistakes later on.

10 to 3 to 1
Apple designers come up with 10 entirely different mock ups of any new
feature. Not, Lopp said, "seven in order to make three look good",
which seems to be a fairly standard practice elsewhere. They'll take
ten, and give themselves room to design without restriction. Later they
whittle that number to three, spend more months on those three and then
finally end up with one strong decision.

Paired Design Meetings
This was really interesting. Every week, the teams have two meetings.
One in which to brainstorm, to forget about constraints and think
freely. As Lopp put it: to "go crazy". Then they also hold a production
meeting, an entirely separate but equally regular meeting which is the
other's antithesis. Here, the designers and engineers are required to
nail everything down, to work out how this crazy idea might actually
work. This process and organization continues throughout the
development of any app, though of course the balance shifts as the app
progresses. But keeping an option for creative thought even at a late
stage is really smart.

Pony Meeting
This refers to a story Lopp told earlier in the session, in which he
described the process of a senior manager outlining what they wanted
from any new application: "I want WYSIWYG... I want it to support major
browsers... I want it to reflect the spirit of the company." Or, as
Lopp put it: "I want a pony!" He added: "Who doesn't? A pony is
gorgeous!" The problem, he said, is that these people are describing
what they think they want. And even if they're misguided, they, as the
ones signing the checks, really cannot be ignored.

The solution, he described, is to take the best ideas from the
paired design meetings and present those to leadership, who might just
decide that some of those ideas are, in fact, their longed-for ponies.
In this way, the ponies morph into deliverables. And the C-suite, who
are quite reasonable in wanting to know what designers are up to, and
absolutely entitled to want to have a say in what's going on, are
involved and included. And that helps to ensure that there are no nasty
mistakes down the line.