# Very Good Post From the Head of Economics at Anthropic > [!question] Lingering Questions > - Levie is repeating someone else's (Anthropic's head of economics) analysis approvingly — does his repost add anything beyond amplification, or is the value purely in surfacing the underlying thread for this vault? > - What demand elasticity would actually be required for the Jevons-paradox effect to offset a given rate of task automation? The post asserts the mechanism without a number. > [!tip] Article Ideas > - None beyond what [[3. Resources/Literature Notes/Lit · Why hasn't AI increased unemployment]] already covers — this post and that source make overlapping arguments from the same intellectual circle. > [!link] Cross Reference > - [[3. Resources/Literature Notes/Lit · Why hasn't AI increased unemployment]] — the fuller version of this same argument, already fully processed in the vault > - [[3. Resources/Permanent Notes/Task Automation Does Not Predict Job Elimination]] — the Permanent note this source already promoted to on 2026-08-11 > [!quote] Vocabulary > - *Jevons paradox* (implied, not named) — the effect where cheaper production of a good increases total consumption enough to offset or exceed the efficiency gain; here applied to labor demand rather than resource consumption ## Bibliographic capsule Aaron Levie, X/Twitter repost (2026-07-24) amplifying an Anthropic labor-economics analysis: AI automates tasks more readily than complete jobs, and cheaper task execution can expand demand enough to preserve or raise labor needs even as individual tasks get cheaper. ## Author's thesis (as I read it) Levie's contribution here is narrow: he surfaces the demand-side half of the automation debate (cheaper production can expand the market rather than just eliminate headcount) as a corrective to purely substitution-focused automation fears. The underlying evidence lives in the source he's amplifying, not in his own commentary. ## Key arguments (≤3) 1. **Task automation and job automation are different units of analysis** — a job survives task-level automation if enough non-automated tasks remain load-bearing. 2. **Cheaper output can expand demand** faster than automation reduces the labor needed per unit, which is a real (if unquantified) countervailing force to simple substitution stories. 3. **The claim is asserted, not measured** in this specific post — no elasticity estimate, no sector breakdown, no time horizon. ## Claim atoms *(Already fully atomized in the existing Permanent note below — not re-deriving.)* ## Permanent candidates - **Task automation does not predict job elimination** — already promoted; see [[3. Resources/Permanent Notes/Task Automation Does Not Predict Job Elimination]]. This source is one of its two origin points (the other being [[3. Resources/Literature Notes/Lit · Why hasn't AI increased unemployment]]). No new Permanent note needed from this row. ## Processing checklist - [x] Confirmed existing Permanent note already covers this source's claim — connected rather than duplicated - [x] Connect into the relevant MOC — [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] ## Links - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — demand-expansion is one of the standing counter-arguments to simple AI-displaces-labor narratives in this MOC's territory - [[3. Resources/Permanent Notes/Task Automation Does Not Predict Job Elimination]] — the claim this source supports, already captured - [[3. Resources/Readwise/Tweets/Very Good Post From the Head of Economics at Anthropic]] — source document