Alistair Cockburn uses the term Shu-Ha-Ri regularly to explain different levels of learning.
The literal translation is approximately Learn/Follow(Shu), Detach(Ha), Transcend(Ri). A number of people I’ve discussed this with equate these levels to the Dreyfus model of skill acquisition. In fact, the Wikipedia article on Dreyfus links back to Shu-Ha-Ri.
In my opinion, this is doing an injustice to both models, and my slightly different interpretation follows.

Driving an example

Take, for example, learning to drive. As I learn to drive, I progress through the Dreyfus model, becoming a novice, then competent, then with a lot of practice, an expert. However, having only learned one way to drive, I would still consider myself to be in the Shu box.

Riding my way to the top

When I learn to ride a motorbike, I can take a number of similar practices from my car experience and apply them to the similar practices in the motorbike world, but I still need to be taught new practices (particularly around observation and hazard avoidance). After a suitable period of tuition, I am now competent in riding a motorbike. I would still consider myself to be a Shu level rider though.

Mixing it up

Now, while driving my car, I suddenly realise that the heightened observation and hazard avoidance techniques, essential while riding, are actually pretty useful while driving. I begin to blend the two skills together. I’m still doing it as taught, but I can pick the practices that are most useful from each and blend them to create something new. This is where I start to become Ha level, taking the most useful practices from a number of techniques and blending them to create something useful in a given context.

Taking it to new heights

Next, I learn to fly a plane, bringing with it new observational practices and new disciplines (e.g. pre-flight checks). Becoming competent here, I can blend some of the practices from flying back into my driving and riding and vice-versa.
Soon I begin to notice a strange feeling. I’m doing all the right things in my practices, but there’s something that seems not quite right, something that could be made better. After some thinking, I identify the missing piece of context and synthesise an appropriate set of techniques for that situation. I have gone beyond the known contexts and identified a new situation, picked or synthesised the appropriate practices for that situation and resolved it. This is Ri level. Having identified this new context, I can teach it to my Shu level students.

But mum, I don’t want to ride a bike

Each of the situations described above are very relevant to me, but that is not to say that you have to ride a bike to progress to Ha or Ri in a car.
Other practices and contexts can be learned on the race track, taking a 4×4 course, rally driving or on a skid pan. Indeed, the beauty of Ha and Ri is that I may bring skills from a completely unrelated area (First Person Shooter, Knitting) and use them to improve my driving techniques.

Softening it up a bit

So, how can we apply this to software? There are a number of practices, techniques and methodologies (e.g. Kanban, Scrum, XP, Lean, Real Options). By the definition explained above, becoming in all of these makes me a Shu level practitioner in each of them.
If I identify a practice from, say, Kanban, that is useful in a context in a Scrum environment, then blending two or more Shu practices takes me to Ha.
When I come across a situation or context where there is no guidance (all of my Shu boxes are empty), then I need to create a new practice or technique. I use my knowledge of similar techniques from a number of methodologies and come up with something that fits the situation.

In Shu-mmary (sorry, summary)

In my interpretation of Shu-Ha-Ri, we have:

• Shu (Learn) – Practice driven. We learn the practices by rote and apply them without understanding of the context.
  Shu practices are our tools, along with an instruction manual on how to use them in the situation they were created for. We can progress through Dreyfus model with each of our tools
• Ha (Blend) – Context driven. Having learned a number of practices, and a number of contexts in which they are appropriate, we can pick or blend a number of different practices for any known context.
  Ha practices are using our Shu tools in situations that weren’t described in their manuals
• Ri (Transcend) – Dealing with unknown context / identifying new practices. Coming across a brand new situation, we are able to identify an appropriate way of handling the situation.
  Ri practices are about creating new tools, either to handle a unique situation, or to handle an existing situation in a better way

I find that separating the Dreyfus model from the Shu-Ha-Ri model allows me to become an expert with each of my Shu tools without feeling that I should be able to apply them to unusual situations.
This is a useful way of thinking (to me at least). Rather than map Shu-Ha-Ri to the Dreyfus model, I choose to think of Shu-Ha-Ri as a skill in identifying contexts and tools.
You might want to discuss this with me some more in the comments 🙂