You divide work into sprints of a week, two weeks, or whatever suits your management style, and you give each sprint a name and a goal (implement search, user registration), then the programmers take stories to go off and make them happen.
Every day your team checks in and tries to unblock one another—if you are working on the tool that sends e-mail and the e-mail server isn’t working, you tell everyone. Then someone else steps up to help, or you stick with that story and do the best you can, but everyone needs to be working toward the sprint goal, trying to release some software. And once the sprint is done, you deliver something that actually, really works and move on to the next thing, slowly bringing a large, complex system into operation.
That’s an ideal case. Done well, it avoids magical thinking (“It will all work when we get everything done and wired together”). It has its critics and can seem to have as many branches (c.f. Scrum, Kanban, and “Agile with Discipline”) as Protestantism.
-Programmers are forever searching for a silver bullet and, worse, they always think they’ve found it. Which is why Frederick Brooks, the most famous of the early software methodologists, wrote a paper called “No Silver Bullet—Essence and Accident in Software Engineering.” He wrote it in 1986. He was very hopeful, back then, that object-oriented programming would help fix things.
+Programmers are forever searching for a silver bullet and, worse, they always think they’ve found it. Which is why Frederick Brooks, the most famous of the early software methodologists, wrote a paper called “No Silver Bullet—Essence and Accident in Software Engineering.” (Link) He wrote it in 1986. He was very hopeful, back then, that object-oriented programming would help fix things.