Podcast Episode: AI And The Economics Of Expertise

Pip: Hemant Pandey has been thinking about what happens after AI gets better at your job than you are — and he’s written it all down, which is either very brave or very on-brand.

Mara: This episode covers two stretches of that argument: first, what AI is already doing to expertise and the economics of teaching, and then where that leads — a possible new model for how human contribution gets recognized and rewarded.

Pip: Let’s start with what happens when the tutor in your pocket outperforms the one you’re paying by the hour.

AI and the Shrinking Scarcity of Expertise

Mara: The central claim here is that expertise has always been valuable because it was scarce — and AI is attacking that scarcity directly.

Pip: The tutoring numbers are the sharpest edge of that argument. The StudentBench study gets cited at length, and one figure in particular is hard to look away from.

Mara: The post puts it plainly: “one AI tutor achieved equivalent measured gains at a reported 918× lower cost per percentage point gained: $0.0052 versus $4.81.”

Pip: So the question stops being whether AI can help the teacher and becomes whether the teacher is still required for the specific function being purchased.

Mara: Right — and the post is careful to frame this narrowly. GRE prep, short interventions, immediate gains. It’s not a universal verdict on teaching. But it’s enough to change the question.

Pip: The calculator didn’t make arithmetic faster — it removed the need to do it manually. The post draws that line explicitly, from calculators through spreadsheets to AI, each one absorbing a layer of cognitive work that used to require a person.

Mara: The companion piece on the economics of instructive teaching extends this into market mechanics. It introduces a price-compression cycle: AI lowers the cost of some instructional tasks, student expectations shift, traditional providers face pressure, and the cycle reinforces itself.

Pip: That piece also notes that price compression in Indian coaching markets was already happening before AI arrived — branded JEE and NEET programs reportedly dropping from roughly ninety thousand to a lakh-and-a-half rupees down toward forty thousand. AI is the second wave, not the first.

Mara: The post’s framing is that instruction is moving from being tied to a teacher’s physical hours, to digitally distributed, to potentially generated by computational systems at scale. The bottleneck moves.

Pip: Though the posts are clear that what remains hard to automate — motivation, accountability, judgment, the ability to notice a student is avoiding hard problems — may actually become more valuable once routine explanation gets cheap.

Mara: The economic principle the posts keep returning to is this: the price of expertise can fall even while the capability of expertise rises. Scarcity and quality are no longer locked together the way they were.

Pip: Which raises the obvious next question — if intelligence becomes abundant, what exactly do humans contribute, and does anyone pay them for it?

Universal Employment and the Knowledge Dividend

Mara: This is where the third piece in the series takes the argument somewhere more speculative, and deliberately so — it opens by calling itself a conceptual framework, not a claim that the mechanisms already exist.

Pip: The core idea is that the industrial model — sell time, receive wages — may increasingly run alongside something different.

Mara: The post frames it directly: “Human contribution → AI improvement → increased value → compensation.” The question it’s asking is whether that second chain can become a real economic pathway.

Pip: What this means in practice is that a teacher’s methodology, a researcher’s edge-case discovery, a domain expert’s correction — these could in principle become inputs to AI systems in ways that generate ongoing value, not just a one-time exchange.

Mara: The post introduces the Soft Brain concept to make this concrete. The valuable asset isn’t the facts an expert holds — it’s the methodology: how they select examples, sequence concepts, diagnose mistakes, decide what comes next.

Pip: Expertise becomes software-like without becoming ordinary software. That’s a genuinely strange thing to sit with.

Mara: The post draws a distinction between Universal Basic Income and what it calls a Knowledge Dividend. UBI asks whether people deserve income as members of society. The Knowledge Dividend asks what should happen when an individual’s contribution creates identifiable economic value through an AI ecosystem — closer to royalties or licensing than to a social transfer.

Pip: The attribution problem is real and the post doesn’t paper over it. Contributions are cumulative and interdependent; isolating what any one person added is technically hard.

Mara: But the economic principle the post wants to establish is clear enough: if intelligence becomes infrastructure built partly through distributed human contributions, some portion of the value should be traceable back to those contributors.

Pip: The YouTube analogy lands well here — mass participation changed who could create and distribute video. The question is whether something similar happens for the creation of machine intelligence itself.

Mara: And the post ends on that note: not a world without work, but one where the production of intelligence becomes more distributed — and where the question shifts from what humans can do despite AI to what humans can contribute to the intelligence AI becomes.

Pip: So the scarcity migrates — from knowledge, to judgment, to direction, to the question of what the machinery should actually be used for.

Mara: That’s the thread connecting all of it. Next time, we’ll see where else that thread leads.