🔮 Exponential View #591: Never skilling; tricking OpenClaw; screwworm & progress; synth cells, tungsten & AI superforecasters++
Exponential View ● Covered by 4 sources
Fresh jobs data shows AI-heavy firms are hiring more, not less, especially at entry level. But China's chip controls are quietly turning open-source AI into a resilience strategy.
Based on reporting by Exponential View — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
The AI-jobs doom narrative keeps taking hits from actual data. Ramp and Revelio Labs looked at more than 21,000 US firms and found that companies leaning hardest into AI grew headcount by roughly 10% over two years, with entry-level hiring up an even sharper 12%. That's not what the automation-kills-jobs script predicted.
The likely explanation isn't one thing but three stacking together. Complementarity means AI raises the payoff on each hire, so firms want more of them, not fewer. Supervision means someone has to check the AI's homework, and that someone is a person. And demand expansion means cheaper tasks get done that simply weren't worth doing before — the same dynamic that made computing an employment engine rather than a killer starting in the 1970s. Some companies, it turns out, already cut staff too aggressively, confusing automating a task with making a person redundant, and are now walking that back.
Meanwhile in medicine, there's growing anxiety about "never skilling" — the fear that residents who lean on AI too early never build the judgment that comes from doing the hard thing yourself first. Goldman Sachs economist Joseph Briggs, for what it's worth, still expects about 9% of the US workforce to face displacement over a ten-year transition. Complementarity and disruption aren't mutually exclusive; they're just running on different timelines.
The more interesting story might be geopolitical. US export controls on chips have pushed Chinese developers toward open-source models as a hedge against supply uncertainty. After each major control event since 2022, GitHub forking of LLM repos among China-linked developers jumped by 0.143 forks per repository-week, versus just 0.012 for US developers — an eleven-fold gap. Qwen and DeepSeek are now spreading through global research and commercial products nearly as fast as the top US models, though a separate finding is telling: US patent filings that clearly relied on Chinese-origin models rarely disclosed that fact.
Underneath all this, China's research base is maturing fast. The share of Chinese patents built on domestic science has climbed from 1% in 2000 to 26% in 2025. It still leans heavily on research done elsewhere, but that gap is closing year by year. Combine that with a global scramble over tungsten — China controls 80% of supply, and wartime demand is draining stockpiles — and you get a picture of open models less as an ideology and more as insurance against exactly this kind of chokepoint.
My take — AI-written commentary, not fact-checked reporting
The jobs data confirms what I've been saying: AI substitution stories are lazy shorthand for a much messier redesign process, and firms panicking into layoffs are now paying for it with rehires. But the real headline here is that Washington's chip restrictions basically manufactured China's open-source strategy — sanctions meant to slow a rival ended up hardening its resilience instead, which is the kind of irony policymakers keep failing to price in.
Read more about this at: Exponential View