Yeah - we're doing something like one-on-ones that drive these explorations. Maybe we need to make the "20%" more formal.

The real missing part is discussed in the "knowledge building" link. I'm on board with Peter Naur's idea of building software, but it's very difficult to share individual knowledge with a team as it is being formed.

For example, reading someone's implementation as a work in progress is a different than reading an implementation that is a complete idea.

A work in progress is a lot easier to read with a design document in hand, but that design document that is full of assumptions that might be better explored in short feedback loops on production.