There's been a lot of talk here in recent posts about project planning and linking to next actions. mdl, I like the point you've made here - it is important to get everything out of your head without predefining steps.

Personally, I use index cards for both NA's and projects: 1 card per NA context and 1 card per project. To link the NA's to the projects, I use a simple notation system. Each project card is numbered, ie. [1], [2], [3] etc. Once a project is completed, I don't worry about re-numbering the cards; just a system of keeping them in order does fine. When I list an NA in a context, I place the number of the project it applies to after it. (Wrap package [2]). I only list those actions I am able to do now in my NA lists.

In my experience, waiting for the weekly review is less effective, especially for smaller personal projects, than is moving onto the next task as soon as possible. I've never fully understood this part of GTD - maybe I'm too impatient to wait for a big review, or my projects demand I review more often. For this reason, I review my completed NAs several times per day, flip to the appropriate project card, and decide the next action. Keeping my project cards in numerical order helps me quickly find the correct project, that is, I don't have to remember how I named a project or where it is filed. Just a stack of numbered index cards.

I also took some of Emory's advice on his promoting technique, and I assign contexts only once an action has been 'promoted' to an NA list. Why try to control that particular part of a project before being in a position to fully understand that part of the project? For me, the outcome of the project is more important than the steps, so as the project evolves, so does my planning. My practice of reviewing my projects as I complete tasks seems to keep me on track.

Anyway, I am sort of rambling here, but I hope this helps. Just another system to consider as we all look for the perfect one.