The original M-shaped model is about building deep expertise in multiple areas, connected by broad interests. But with AI, this reframes: depth is no longer about execution. It’s about understanding enough to see patterns and direct the AI.

So what do I actually do? AI handles the technical depth, the execution, the details. I handle planning, context, spotting connections between distant things. Far Transfer (seeing a pattern in one domain and applying it to another) stays with me. I haven’t seen AI do this well yet.

The 80% rule also transforms. It used to take hard work to reach 80% fluency in a field. Now with AI, I can operate at 80% without actually being there (AI fills the gaps). But I need minimum understanding, otherwise I won’t notice when it’s wrong.

I don’t write code anymore, but I understand it. Review is harder than writing (you have to understand what someone else intended, not just what you meant). Daily code review from AI = daily practice in understanding. The trade-off is real though: not writing means some muscle memory fades. I’m betting that review keeps enough intuition alive.

GTD and Zettelkasten work as multipliers here. GTD tells me what work needs doing, Zettelkasten tells me how to structure learning, AI handles the execution. I can go deep in more areas than before because 1.1.1.1 the execution cost dropped.