A year ago, the men building artificial intelligence were predicting the end of work as we know it. Today, these same individuals have shifted their narrative. The data did not prove them right.

In May 2025, Dario Amodei —CEO of Anthropic— maintained that AI could eliminate 50% of entry-level white-collar jobs. He projected a U.S. unemployment rate of between 10% and 20% within a one-to-five-year timeframe. One year later, that rate remains at 4.3%. Sam Altman, at the helm of OpenAI, anticipated that AI agents would "join the workforce" during 2025. That did not happen either.

Both have moderated their tone. Amodei now speaks of Jevons' paradox —the idea that technological efficiency increases consumption rather than reducing it. Altman migrated from "AI replaces" to "AI transforms." The language became more cautious just as their companies' valuations approach the stock markets. Anthropic projects its IPO for late 2026. OpenAI is following the same path.

This is no minor coincidence.

Andrew Ng pointed it out this week without much beating around the bush: the jobpocalypse is a narrative serving concrete interests. AI companies anchor their pricing to the salaries of the employees they are supposedly going to replace. Companies that laid off staff post-pandemic —when capital was cheap and hiring was excessive— found in AI a more elegant explanation than admitting overhiring. And the more disruptive a technology sounds, the higher its market valuation.

Ng is right, but he is missing one turn.

The replacement narrative also solves an audience problem. Developers make up 0.6% of the population. Enterprise software buyers represent the remaining 99.4%. Telling a CEO "replace your devs" is more powerful than telling them "help your engineers be a little more productive." Drama sells. Gradualism does not.

The data, meanwhile, tells a different story.

In the first quarter of 2026, U.S. tech companies announced 81,747 layoffs, the highest number since 2024. 47.9% of those cuts were attributed to AI and automation by the companies themselves. Amazon eliminated 16,000 corporate positions. Oracle, between 10,000 and 30,000. Meta, 8,000. That same quarter, AWS —Amazon's cloud division— grew by 24%, the fastest pace in 13 quarters. Marc Benioff, CEO of Salesforce, summarized it without euphemisms after eliminating 4,000 customer service roles: "I need fewer heads." However, Anthropic's own Economic Index —published by the company with the most to gain from the replacement narrative— shows that 52% of Claude's usage corresponds to supporting tasks, not substitution. The Yale Budget Lab found no statistically significant differences in employment between occupations exposed to AI and those not exposed since the launch of ChatGPT through March 2026. Software engineer hiring, the sector most impacted by code agents, remains steady.

The jobpocalypse is a narrative without evidence.

This does not mean that nothing is changing. The transformation of work is real, profound, and, for many, difficult. Tasks are disappearing. Roles are being redesigned. Skills that were highly valued are now worth less. This deserves attention, public policy, and honesty. What it does not deserve is the disguise of a civilizational collapse that some put on top of it to sell technology, justify layoffs, or inflate a valuation before going public.

Rocbird also sells AI services. We state this because the honesty we demand from Anthropic and OpenAI applies to us as well. The difference we propose is to work with data rather than predictions, and with evidence rather than fear. Language models do not read the future. It has been proven that neither do their creators.