Divya van Mahajan

Is AI Doom Going Out of Style?

ai-daily-brief-podcast

Is AI Doom Going Out of Style?

Overview

This episode of the AI Daily Brief podcast examines emerging signals of a potential “vibe shift” in public discourse around AI and its economic impact — specifically, whether the dominant narrative of AI-driven mass unemployment and doom is beginning to soften. The host argues that this shift is visible simultaneously in intellectual commentary (particularly among economists and political commentators) and in market data, making it more likely to represent a genuine narrative evolution rather than a temporary blip. No single speaker affiliation is stated beyond the show itself.

Source: AI Daily Brief — Episode: “Is AI Doom Going Out of Style?” (recorded ~May 4, 2026) (No YouTube URL was provided for this episode.)

Prerequisites

  • Basic familiarity with the current AI industry landscape (LLMs, AI coding tools, agentic AI)
  • Understanding of macroeconomic concepts: unemployment rates, labor market dynamics, CapEx
  • Awareness of the public debate around AI and job displacement
  • Familiarity with Jevons’ Paradox (efficiency gains increase, rather than decrease, overall resource consumption)
  • Basic understanding of SaaS business models (seat-based vs. usage-based pricing, ARR)
  • Knowledge of major AI companies: OpenAI, Anthropic, Microsoft, Google

Main Points

1. Two Competing Narratives Frame the Debate

  • A New York Times essay by Jasmine Sun (“Silicon Valley Is Bracing for a Permanent Underclass”) represents the doomer pole, sourcing its pessimism from AI builders in San Francisco and Silicon Valley.
  • The host argues that AI builders are structurally poor predictors of AI’s societal impact: they may be overestimating their technology’s reach, have financial incentives (IPOs) to hype disruption, and lack exposure to economics, labor markets, and non-startup contexts.
  • The doomer narrative is not simply wrong, but it is being built on a narrow evidentiary base from a self-interested group.

2. Ezra Klein’s Reframing: Economists Are Skeptical of Mass Joblessness

  • Ezra Klein’s New York Times piece, “Why the AI Job Apocalypse Probably Won’t Happen,” is notable precisely because Klein is not an AI accelerationist — his voice carries different weight than a Marc Andreessen making the same argument.
  • Klein draws on University of Chicago economist Alex Emas’s essay (“What Will Be Scarce”) and the concept of Jevons’ Paradox: when technology reduces the cost of doing something, demand for that thing typically expands rather than collapses.
  • Historical precedent from ASU professor Eldar Maximoff: in every major occupation group that adopted computers heavily, employment grew faster than in groups that did not. Computers eliminated tasks but expanded overall occupational demand.
  • Klein’s personal example: his podcast team has grown, not shrunk, with productivity tools — more capability creates more ambition, not more leisure.
  • Klein’s key line: “Every enthusiastic AI adopter I know is working harder than ever because there is more they can do.”

3. Klein’s Nuanced Warning: The Middle-Displacement Scenario

  • Klein does not argue everything will be fine. His concern is that partial displacement may be harder to manage politically and socially than mass displacement.
  • A world where AI displaces 8 million workers may be worse than one displacing 80 million — mass unemployment would force wholesale economic restructuring; limited displacement is easier to ignore.
  • The China trade shock (approximately 2 million jobs lost) was small in macro terms but devastating to specific communities, with very little government response.
  • The risk is not total job loss but rather concentrated, inadequately addressed displacement in specific job categories.

4. Macro Data Does Not (Yet) Match the Doom Narrative

  • Unemployment rate in March 2026: 4.3% — virtually unchanged from 4.4% in March 2020, before the AI boom.
  • Average hourly earnings are stable.
  • Software engineering job postings are up 18% from May 2025, now at their highest point since November 2023 (Federal Reserve data).
  • Citadel Securities data cited by Sequoia’s Konstantin Buehler: demand for software engineers — the most AI-exposed occupation — is accelerating, not declining.
  • Investor Anthony Pompliano updated his prior belief: new college grad hiring is up 5.6% year-over-year; unemployment for degree-holding 20–24-year-olds fell from ~9% to ~5%.
  • LinkedIn/Wall Street Journal analysis: AI created 640,000 jobs in the U.S. between 2023 and 2025.

5. Elastic vs. Inelastic Demand — Where Jevons’ Paradox Applies

  • Murza Ahmed (Emergence AI / Russell AI Labs) frames the key distinction: elastic demand (software development, sales outreach, legal discovery, security monitoring) expands when cost falls; inelastic demand (payroll, month-end close, routine compliance) does not.
  • Code is described as “digital brick”: cheaper bricks mean more building, not fewer builders.
  • Alex Emas’s essay (the underlying economic framework) argues that savings from AI in one sector flow into high-elasticity sectors — particularly relational, bespoke, human-mediated experiences.

6. The Entrepreneurship Signal

  • Greg Eisenberg (Startup Ideas podcast) argues we are approaching the largest explosion of entrepreneurship in human history — not because job loss won’t happen, but because displaced workers and new entrants will use AI to start businesses that were previously too costly or slow to launch.
  • Stripe Atlas data (Patrick Collison, May 1, 2026): 100,000 all-time incorporations; Q1 was 130% up year-over-year.
  • Derek Thompson: Stripe data shows both startup incorporations way up and AI startups seeing faster-than-historically-normal revenue growth.
  • Current observation: “AI agents are better at creating firms than destroying jobs.”

7. The AI Bubble Narrative Is Softening in Markets

  • The Atlantic (Roger Karma): six months ago, AI looked bubbly — massive CapEx with no clear path to profitability. That framing is shifting.
  • The pivot: from seats to tokens. Earlier skepticism was based on counting $20–$30/month subscriptions and questioning whether that justified trillion-dollar infrastructure. Agentic AI changes this: a single user could consume hundreds or thousands of dollars of tokens per month with no upper ceiling tied to a seat.
  • Anthropic ARR: SemiAnalysis (considered well-sourced) reports ARR exploded from $9 billion to over $44 billion in 2026, approximately doubling every six weeks. Anthropic is reportedly adding ~$96 million in ARR per day.
    • For context: AWS took 13 years to reach $35B ARR; Salesforce took 20+ years to pass $20B.
  • Anthropic inference margins reportedly improved from 38% to 70% gross margin year-over-year.
  • Morgan Stanley raised CapEx forecasts for five hyperscalers to $805 billion in 2026, with 2027 forecasts at $1.1 trillion.
  • Demand backlog from hyperscaler customers (~$1.3 trillion) is growing faster than CapEx spend — arguing against a bubble dynamic.
  • David Sacks estimates AI CapEx represents a 2.5% tailwind to U.S. GDP in 2026, rising to over 3% in 2027 — and this understates total economic impact since it excludes productivity gains from token consumption.

8. The SaaSocalypse Narrative Is Also Shifting

  • Atlassian’s earnings (stock up ~30%): revenue grew 32% year-over-year.
  • Key product: Rovo (AI search/knowledge tool). Customers using Rovo were growing their own ARR at twice the pace of non-users.
  • Atlassian’s structural advantage: 20 years of captured structured data in Jira and Confluence enables graph lookups instead of token-hungry RAG (retrieval-augmented generation), making their AI tool more efficient than generic enterprise AI.
  • Analyst conclusion: seat compression from AI is not visible in Atlassian’s case; seat-based pricing remains rational given their data moat.
  • Market interpretation: fears that AI agents would replace SaaS platforms are, at least in some cases, proving premature.

9. Messaging Shift from AI Companies Themselves

  • Sam Altman (OpenAI, May 1, 2026): “Jobs doomerism is likely long-term wrong.” Frames OpenAI’s goal as augmenting people, not replacing them; expresses hope for a future of fulfilling work and optional labor.
  • The host notes this is a significant rhetorical pivot: for years, replacing humanity was an explicit stated goal of OpenAI’s leadership.
  • Noah Smith characterizes it as “a huge messaging pivot.”
  • The host’s critique: Altman’s framing still underestimates how hard economic diffusion is and how capable market economies are at matching new capital and labor with new demand — but the directional change is positive.

10. Caveats: The Vibe Shift Is Faint and Contested

  • A concurrent New York Times piece reported that worry about AI is the one issue uniting Democrats and Republicans — public sentiment remains negative.
  • The “abundance” intellectual framework (Klein, Derek Thompson) is actively rejected by significant portions of the political left.
  • Data center opposition from communities remains a live issue; construction trade unions are now allying with tech companies to counter opposition by highlighting construction jobs created.
  • The host cautions that he may be pattern-matching a theme from limited data points (one essay, some market signals), and the dominant doomer narrative has not collapsed.

Key Concepts

  • AI Vibe Shift: The emerging, tentative change in dominant public and market discourse from AI-as-catastrophe toward more nuanced or even optimistic economic framings.
  • Jevons’ Paradox: The economic phenomenon whereby increased efficiency in resource use leads to greater total consumption of that resource, not less — applied here to AI making labor more productive and thereby increasing demand for labor.
  • Elastic vs. Inelastic Demand: Elastic demand expands when the cost of a good or service falls (e.g., software development); inelastic demand does not meaningfully expand with cost reduction (e.g., routine compliance filing).
  • Seats to Tokens: A framing shift in how AI revenue is understood — from fixed per-user subscription pricing to consumption-based token pricing, which has no inherent ceiling per user.
  • Agentic AI / Agent Era: AI systems that autonomously execute multi-step tasks (e.g., Claude Code, OpenAI Codex), driving token consumption far beyond simple chat interactions.
  • RAG (Retrieval-Augmented Generation): A technique where AI retrieves external documents to provide context for responses; computationally expensive in tokens compared to structured knowledge graph lookups.
  • SaaSocalypse: The feared scenario in which AI agents replace enterprise SaaS platforms by replicating their functionality on demand.
  • ARR (Annual Recurring Revenue): A standard SaaS metric measuring annualized subscription or recurring revenue.
  • CapEx (Capital Expenditure): Large-scale investment spending by hyperscalers (Google, Microsoft, Amazon, Meta, etc.) in data centers and AI infrastructure.
  • Hyperscalers: The largest cloud computing companies (referred to in the episode as the “Mag7” or “five hyperscalers”) whose infrastructure investments dominate AI compute supply.
  • Stripe Atlas: Stripe’s service for incorporating startups, used here as a proxy for global entrepreneurship activity.
  • The Relational Sector: Alex Emas’s term for industries providing bespoke, human-mediated, high-touch experiences — posited as a primary beneficiary of productivity savings freed up by AI elsewhere.

Summary

The episode argues that a tentative but potentially significant “vibe shift” is underway in how AI’s economic consequences are being discussed. The dominant doom narrative — AI will cause mass unemployment — is being challenged from two simultaneous directions: intellectually, by economists and prominent commentators like Ezra Klein drawing on Jevons’ Paradox and historical labor market data to argue that AI is more likely to expand demand for work than eliminate it; and empirically, by labor market and market data showing software engineering job postings rising, unemployment stable, startup formation accelerating, and AI company revenues growing at historically unprecedented rates. The episode does not dismiss the possibility of real disruption — Klein’s warning about concentrated, inadequately addressed job displacement in specific communities is treated as the most important and underappreciated risk — but contends that the extreme doomer framing systematically overweights the perspective of Silicon Valley insiders while underweighting the structural adaptability of market economies. The host closes by expressing cautious optimism that this moment of narrative softening creates space for a more productive conversation: not doom, not utopia, but practical engagement with how to manage and maximize the transition ahead.