Andrej Karpathy—early OpenAI researcher, former head of AI at Tesla—recently said he’s never felt this behind as a programmer. The programming profession, he wrote, is being “dramatically refactored.” The contributions from the programmer are becoming “increasingly sparse.”
Read that again. Someone who literally helped build these systems feels the ground shifting underneath decades of accumulated expertise. If that’s true for him, what does it mean for everyone else?
Here’s the part that matters for organizational leaders: Karpathy also said he senses he could be “10x more powerful” if he figured out how to properly work with what’s emerging. The gap isn’t between practitioners and obsolescence. The gap is between where people are now and where they could be if they learned to work differently. But “differently” isn’t a training module. It’s a new skill tree—and right now, most organizations haven’t defined what that tree looks like.
The decision at the heart of this briefing
We’re stuck in a world where technical teams have their skill tree and non-technical teams have theirs. Engineering levels up through code review, system design, debugging. Legal levels up through contract drafting, negotiation, regulatory knowledge. Finance levels up through modeling, analysis, forecasting. All of these skill trees assume manual problem solving: humans do the cognitive work, tools make it faster or more accurate.
That assumption is breaking. The skill trees are going to merge—not because AI makes everyone “technical,” but because the skills that matter when AI handles generation are the same across functions: specifying intent clearly, maintaining authority over outputs you didn’t fully create, building workflows that don’t depend on individual heroics, and creating systems that improve over time instead of resetting with every project.
The lawyer building a contract review workflow and the engineer building a debugging assistant are now climbing the same tree. Different artifacts, different tools, same hierarchy of capabilities. Leaders who define that tree give their teams something to climb toward. Leaders who don’t leave everyone stuck—leveling up on skill trees built for a world that’s disappearing.
The decision this briefing asks you to make: Are you going to keep separate skill trees for technical and non-technical roles, or define a unified AI problem-solving tree and re-level your organization around it? That’s not a philosophical question. It’s a budget question, a hiring question, a “what does competence mean here” question. Everything that follows is designed to make that decision concrete.
This briefing covers:
Why the old skill trees are breaking—the shift from manual problem solving to orchestrating AI-assisted work, and why it affects every function
The new skill tree—four levels of capability that apply across legal, finance, engineering, operations, and every other knowledge function
The fork at the base—tool-mode versus infrastructure-mode as the first decision on the new tree
What leaders need to define—how to lay out the skill tree so teams can actually level up
How to get started—building proof that the new tree works in your organization
