You're exactly right about software being exploratory, but I think the whole notion of fixed, up-front requirements is part of the problem. In the lean analysi

You're exactly right about software being exploratory, but I think the whole notion of fixed, up-front requirements is part of the problem. In the lean analysis, the problem isn't changing requirements, which is a natural consequence of changing circumstances and people learning over time. The problem is building up a large inventory of poorly-tested, underinformed plans and then doing a lot of work based on them without taking advantage of the opportunities to learn. Especially here on HN, we know that startups don't succeed because they come up with a fixed initial plan and then spend years marching to it. Instead we have all set of techniques for releasing early and often so we can see what really works for users. A key competitive advantage for startups is how their fast OODA loops allow them to run rings around larger companies that slowly drift into being very plan-based.