He's an industry leader. In my limited experience, there is very little I can add to this. I can't say it any better. I can only sit at the back of the room and look at everyone else to see if they are getting it as clearly as I am.
For those of you who haven't seen his work before, here are my favorite excerpts to entice you to read the full article:
Starting a project- "You don't need anywhere near the level of documentation, or detail for that matter, that you've been led to believe over the years. For example, although you'll likely be asked for an initial estimate, you don't need to spend weeks or months doing detailed requirements, analysis, and design so that you have the input into some form of complicated estimating process. A ballpark estimate will often do, and frankly it is just as accurate as anything you'd produce counting function points or following COCOMO II."
Should you do a project? "An important consideration during Cycle 0 is the economic, technical, operational, and political feasibility of the project. So we ask ourselves if the system will pay for itself; if you have the ability to build the system (or to manage the project if you're outsourcing the effort); if you can keep the system running once it's built; and whether your organization can tolerate the successful delivery of this system."
Cost? Deal with it- "Agile or not, someone is going to ask the questions: "How much is this going to cost?", "How long is this going to take?", and "What am I going to get?". The real answers, respectively, are "As much as you're willing to spend," "As long as it takes," and "Whatever you tell us that you want," but few organizations are willing to tolerate this level of honesty and instead ask for "exact" answers. Striving for exact answers typically results in both time and money being wasted with little improvement in accuracy to show for it. I believe that it is time to admit that we can't predict the future accurately and cast off the shackles of traditional thinking."