Is AI Doom Going Out of Style?

May 5, 2026 · Episode Links & Takeaways

Author

AI Daily Brief
May 04, 2026

Please help us out by taking a few minutes to fill out this month’s AI Pulse Survey.

MAIN STORY

Is AI Doom Going Out of Style?

Something is shifting — faintly, but meaningfully — in how the broader culture is talking about AI and jobs. The signals are coming from two directions at once: the chattering class is producing more nuanced takes that push back on mass unemployment narratives, and markets are finally validating the agentic revenue thesis in real time. When a narrative shift happens on both sides simultaneously, it's worth paying attention to.

HAS THE AI VIBESHIFT ARRIVED?

The Two Poles: Jasmine Sun vs. Ezra Klein
The mirror held up to Silicon Valley — and the pushback that followed.
The backdrop to this week's vibe shift is a Jasmine Sun essay in the New York Times arguing Silicon Valley is bracing for a permanent underclass — a well-sourced piece reflecting how AI builders themselves talk about job displacement. The problem isn't the sourcing; it's that the people closest to AI have both incentives (IPOs looming) and blind spots (they're not economists) that make them structurally unreliable narrators about labor market impacts. The more important piece this week was Ezra Klein's response, which matters precisely because it doesn't come from the usual accelerationist voices.

Alex Imas & Ezra Klein: The Economics Case
Jevons' Paradox reaches a mainstream audience.
Ezra's piece builds on Alex Imas's "What Will Be Scarce?" essay, bringing demand elasticity and Jevons' Paradox to a mainstream readership. The core insight: in every major occupation group that adopted computers heavily, employment grew faster than in groups that didn't. Computers eliminated specific tasks, but the resulting cost reductions created so much new demand that occupations expanded overall. Ezra's personal example is telling — his podcast team has grown, not shrunk, because AI brings him more to absorb and consider, so he chooses to do more challenging work. His conclusion isn't Pollyannish: the transition can be brutally painful for specific workers and communities, and we've historically been terrible at helping those people. A world where AI displaces 8 million workers may actually be harder to navigate politically than one where it displaces 80 million.

Alex Imas (Substack) What Will Be Scarce?
Alex Imas (X) Companion tweet

The Data: Software Engineering Jobs Are Booming
The most AI-exposed occupation is accelerating, not collapsing.
The macro data isn't matching the doom narrative. Unemployment sits at 4.3% — nearly identical to March 2020 before the pandemic. Software engineering job postings are up 18% from their inflection point in May last year, and have hit their highest level since November 2023. Sequoia partner Konstantine Buhler flags the Citadel Securities analysis showing the acceleration has continued even after the "transitional integration" explanation was offered as a rebuttal. Murtaza Ahmed at Russell AI Labs goes further, arguing that AI driving demand for software engineers is now actually the consensus view in tech circles: code is digital brick, and when bricks get cheaper, you build more — you don't hire fewer builders. The WSJ reported that AI created 640,000 new jobs between 2023 and 2025 per LinkedIn job posting data. Anthony Pompliano, who publicly changed his mind on AI job displacement this week, noted that new college grad hiring is up 5.6% over the last 12 months and unemployment for 20-to-24-year-old college graduates has fallen from nearly 9% to almost 5%.

The Atlantic: Markets Wake Up to the Agentic Era
Revenues are finally catching up to the hype — because of Claude Code.
The Atlantic's Rogé Karma published a piece this week explicitly revisiting the AI bubble thesis. Six months ago the sector looked bubbly, with hundreds of billions flowing into data centers with no clear path to profitability. Today, the concern has inverted: we may not build enough to satisfy demand. The cause of the turnaround, he writes, can be summarized in two words: Claude Code. The shift from seats to tokens is the key: in the agentic era there's effectively no ceiling on consumption, and a single developer doesn't represent a $20 seat — they represent hundreds or thousands of dollars a month flowing to the token sellers. Anthropic, per Semianalysis, has gone from $9B ARR at the start of the year to over $44B today, with inference margins expanding from 38% to over 70% over the same period. Analyst Meng Li's back-of-napkin math: Anthropic is adding roughly $96M in ARR per day. AWS took 13 years to reach $35B in annual revenue. Salesforce took over 20 years to pass $20B.

CapEx Reframed: Backlog > Spend
When demand exceeds supply, CapEx isn't a bubble — it's a queue.
Morgan Stanley has raised its hyperscaler CapEx forecast to $805B for this year and $1.1T for next. The standard bear case is that this looks like dotcom fiber overbuilding. Andreas Steno Larsen pushes back with the key data point: the backlog of customer demand for additional capacity is diverging upward from CapEx spend, not converging. Mag-7 companies spent over $400B in CapEx in Q1, but their reported and projected backlog sits around $1.3T. David Sacks contextualizes the macro angle: AI CapEx is already a 2.5% tailwind to GDP this year per Morgan Stanley, and that number understates it because it only counts the five hyperscalers and doesn't capture the economic activity happening inside the token factories themselves.

Atlassian: The SaaS Apocalypse That Wasn't
Customers want to buy AI, not build it themselves.
Atlassian stock was up almost 30% on Friday after reporting revenue growth of 32% year over year, up from 23% last quarter. The bigger story was rapid adoption of their AI search tool Rovo: customers using it are growing their own ARR at twice the pace of those who aren't. CEO Mike Cannon-Brookes explained the token efficiency advantage — Rovo does a knowledge graph lookup rather than token-hungry RAG search, because Atlassian and their customers have spent 20 years building structured relationships between work items, people, code, and knowledge. The market read this correctly: structured databases are a real moat. Customers aren't vibecoding their own Jira replacement. Jason Lemkin asks the right question: were public software companies oversold?

Blue Collar Alliance
Building trades unions are now intertwined with Big Tech.
AP reports that construction unions are becoming unlikely allies for AI companies in the data center permitting battles. Rob Bear of the Pennsylvania Building and Construction Trades Council: communities should stop saying "no" and start asking what they need — school funding, project improvements — because data centers do create a lot of construction jobs. The unstated implication: it's an indictment of how poorly tech companies have managed these rollouts that community opposition has reached this level. There are so many ways to make data centers value-accretive to surrounding communities at a fraction of total project cost.

OpenAI's Messaging Pivot
Sam Altman finally says the thing he should have been saying for years.
On May 1st, Sam Altman tweeted that OpenAI wants to build tools to augment and elevate people, not entities to replace them. He added that jobs doomerism is "likely long-term wrong." On a recent podcast, he echoed the Ezra Klein theme: someone told him GPT-5.5 and Codex can accomplish in an hour what would have taken weeks two years ago, "and I've never been busier in my life." Noah Smith, who has been loudly tracking the AI narrative, called it a "huge messaging pivot" — for many years, replacing humanity was the explicit stated goal.

Is the Vibeshift Overstated?
Probably, at least a little — and it's worth saying so clearly.
This whole episode is built on a few market signals and one essay by a prominent left-leaning commentator, which is a thin foundation for declaring a narrative shift. The New York Times published a piece Sunday finding that the one issue uniting Democrats and Republicans right now is worries about AI — and polls still show people are extremely negative. Large parts of the left reject the abundance framework out of hand, which happens to be the intellectual scaffolding behind both Klein and Derek Thompson's optimism. Even so, there's real value in the collective foot being taken off the gas of AI doomerism for just a moment — not to replace it with utopia, but to open space for a more grounded conversation about how to actually adapt to the change that's here.</p>