Every Automation Wave Has a Redistribution Pattern — AI Is No Different

1/ The spinning jenny. The assembly line. The spreadsheet. Every automation wave follows a pattern: destroy some jobs, create others, and redistribute wealth — usually upward. AI is no different. Here's the history. 🧵

2/ TEXTILES, 1760s: The spinning jenny multiplied a worker's output 8x. Hand spinners were ruined. Mill owners got rich. Eventually factory workers organized, won protections, and shared the gains. It took 60 years.

3/ ASSEMBLY LINE, 1913: Ford's line cut Model T production from 12 hours to 93 minutes. Ford doubled wages to $5/day — not from generosity, but because turnover was 370%. Workers had leverage. They used it.

4/ OFFICE AUTOMATION, 1980s: The spreadsheet eliminated thousands of bookkeeping jobs. But it created financial analysis as a mass profession. The catch: the new jobs paid better, but required different skills. Not everyone made the jump.

5/ ATMs, 1970s-2000s: Banks installed ATMs and the number of bank tellers... increased. Cheaper branches meant more branches. But the job changed: less cash handling, more selling financial products. Automation reshapes, not just replaces.

6/ Daron Acemoglu calls the worst version "so-so automation": tech that replaces workers without meaningful productivity gains. Self-checkout kiosks don't make shopping faster. They shift labor from paid workers to unpaid customers.

7/ The crucial variable isn't the technology — it's the power balance. When workers had unions, automation gains were shared (postwar boom). When they didn't, gains went to capital (1980s onward). The technology is neutral; the institutions aren't.

8/ Erik Brynjolfsson documented "the great decoupling": from 1948-1973, productivity and wages rose together. After 1973, productivity kept climbing while median wages flatlined. The gains were redistributed — to the top.

9/ AI follows the same fork. Path A: AI augments workers, raises productivity, gains are shared through bargaining and policy. Path B: AI replaces workers, gains go to owners, and displaced workers compete for worse jobs.

10/ Early evidence is mixed. GitHub Copilot makes junior developers 55% faster. But AI coding tools also let companies hire fewer developers. Augmentation and replacement aren't opposites — they're a sequence.

11/ History's clearest lesson: the adjustment period is brutal, and it falls hardest on those with the least power. The handloom weavers weren't wrong that machines threatened them. They were right. They just lost.

12/ The question isn't "Will AI automate jobs?" It already is. The question is whether we repeat the pattern — decades of pain before policy catches up — or learn from 250 years of evidence and act faster this time. /end

---

LinkedIn version:

Every automation wave follows the same pattern: destroy some jobs, create others, redistribute wealth — usually upward. AI is no different, and 250 years of history tells us exactly what to watch for.

The spinning jenny multiplied a spinner's output 8x. Hand spinners were ruined; mill owners prospered. It took 60 years of organizing before workers shared the gains. Ford's assembly line cut production from 12 hours to 93 minutes, but only doubled wages when 370% turnover forced his hand. ATMs didn't eliminate bank tellers — they made branches cheaper, so banks opened more — but the job transformed entirely.

Daron Acemoglu identifies the worst scenario: "so-so automation" that replaces workers without real productivity gains. Self-checkout doesn't make shopping faster. It shifts labor from paid workers to unpaid customers.

The crucial variable across every wave isn't the technology — it's the power balance. Erik Brynjolfsson documented "the great decoupling": from 1948-1973, productivity and wages rose together. After 1973, productivity kept climbing while median wages flatlined. What changed wasn't the pace of technology. It was the decline of worker bargaining power.

AI sits at the same fork. Path A: augment workers, share gains through bargaining and policy. Path B: replace workers, concentrate gains at the top. Early evidence is mixed — GitHub Copilot makes developers 55% faster, but companies are also hiring fewer of them.

History's clearest lesson: adjustment periods are brutal and fall hardest on those with the least power. The question is whether we repeat the pattern or learn from it.