▶ YouTube Summaries

← All summaries
The PrimeTime#▶ Watch

CEOs walk back the AI jobs apocalypse

2026-05-29 Fri ⏱ 13 min @ThePrimeTimeagen

Prime argues the narrative is shifting: the executives who spent two years predicting AI would gut entry-level white-collar jobs are now quietly reversing course, and the promised 10–100x productivity gains aren't showing up anywhere except the model labs' own revenue.

The walk-back

  • Sam Altman (Commonwealth Bank conference, Sydney) now says he doesn't expect the "jobs apocalypse" some companies in his space "advocate" for — a thinly veiled jab at Anthropic/Dario Amodei. Admits his intuitions were "just off" and that people genuinely value human interaction.
  • Goldman Sachs CEO David Solomon: AI won't eliminate 25% of jobs; people redirect freed time to more productive work, and the doom scenario isn't in the data.
  • Contrast with the prior era: Amodei warning of 10–20% unemployment, Altman's "Death Star" tweet — messaging Prime found confusing and self-defeating.

The productivity that isn't there

  • Uber COO: heavy AI spend is getting harder to justify. Stats like "25% of commits were AI-driven" or rising token usage don't trace to more shipped consumer features. The link "is not there yet." Uber burned its full operating-year AI budget in four months.
  • Linear CEO: we hear about 10–100x gains in engineering, but outside model labs there's no matching 10–100x revenue growth or quality jump. "The only people making money are the people selling shovels."

Prime's own take

  • Mocks the "just one more harness, bro" mentality — compares endless agent-harness tuning to the old Neovim-config-tweaking meme. Sharpening the sword instead of cutting.
  • Where AI helps him: exploration is cheap. He describes an idea, has the model build several mock versions, compares interfaces, iterates (e.g. animations in an immediate-mode UI). Speeds up understanding, not necessarily implementation.
  • For deep, integrated products (going 30→60, not 0→1), leaning on AI often makes things more brittle; you still must step back and work the problem yourself. Good decisions feel harder now because you move so fast.
  • Rejects both doom and hype. Still calls it magic — broken-English prompts producing roughly what you describe — but insists competency and learning to program well remain important.