Tan’s sharpest claim is that a skill file is not a document but “a piece of your cognition, how you do the thing, extracted from your head, written down, and executable.” He runs it through a deliberately ordinary fictional case: a support engineer, Maya, who over two years teaches her agents 40 skills — triaging a P0 at 2am, de-escalating a customer about to churn, writing a postmortem that actually prevents the next incident. Same forty files, two opposite futures. In her repo, she changes jobs and arrives on day one with years of compounded judgment on tap. In the company’s repo under the company’s IT policy, she leaves with nothing while the company keeps running her judgment without her, her name not even in the commit history: “She didn’t have a career, she had an extraction.”
The historical frame is what gives it teeth. Craftsmen owned their tools and that is what made them free; the factory broke it, and the loom belonged to the mill. Knowledge workers assumed they were safe because their tools lived in their heads where nobody could confiscate them — and Tan’s argument is that skill files end that exemption, because for the first time cognition can be extracted, stored, versioned and owned. This is the personal-sovereignty counterpart to Domain-specific skill libraries are the real agent moat, not core infrastructure, which makes the same argument at firm level; read together they say the moat is real and the open question is only who holds it. It is why Personal AGI is an asset you build, not a subscription you rent is framed as ownership rather than convenience, and it gives a motive for the discipline in Compound engineering makes each unit of work improve all future work — every skillified task is an entry in an asset register, and the register has an owner either way.