Anthropic Merges 30,000 Changes a Day, and It Shows
Anthropic now merges over 30,000 code changes a day, up from around 2,500 in May. It says so in an engineering post…
Every economy wants higher productivity. When people get a few percent more done each year, the whole economy grows with it. But I think AI could make us so productive that it does the opposite. If one person can suddenly do the work of 20, the economy could shrink instead of grow, or even collapse.
Normal productivity growth is slow. In the US, it has averaged 2.1% a year since 1947, according to the Bureau of Labor Statistics. That's the kind of growth you want. If the people at a company get a few percent more productive, the company grows a bit. Or it lets a handful of people go, and other companies are happy to hire them, because everyone needs people.
Even huge changes worked out when they were slow. In 1900, 41% of American workers worked in agriculture. By 2000 it was 1.9%, according to the USDA. That took a hundred years, and the farm workers and their kids had time to move into factories and offices.
AI isn't a few percent. If it does what everybody thinks it does, one person will soon do the work of 10 or 20 people, maybe even 100. That goes for software development, and for almost everything else people do in front of a computer.
Dario Amodei, the CEO of Anthropic, described one version of this future to Axios last year. In his words, "the economy grows at 10% a year" while "20% of people don't have jobs." I don't see how that works. People without a job can't buy much.
Say a bank has 5,000 people in its head office. They check loan applications and answer customer emails, and almost everything they do happens on a computer. With AI, 500 people can do the same, so the bank lets 4,500 of them go.
With normal productivity growth, other companies would hire them. But every other bank and every insurance company is doing the same thing at the same time. Nobody needs 4,500 more office workers, so they stay unemployed and stop spending money.
The bank also stops spending money on the people it let go. It used to rent 40 floors of offices, and now four are enough. It paid for 5,000 software licenses that are billed per person, and now it pays for 500. So the landlord and the software company lose most of the money they made from this bank. The canteen and the coffee shop downstairs lose most of their customers.
Then there are the 4,500 people. They don't buy a new car this year, and they eat out less. Some of them can't pay their mortgage anymore. In the US, consumer spending is about two thirds of the whole economy. When millions of people stop spending at the same time, every business that sells to them feels it.
And the bank's customers are these same people. They are the ones who take out mortgages and car loans. The bank let 4,500 people go to make more money, and now it has fewer customers, because every other company did the same.
In Will AI Grow the Pie or Just Hand It to Fewer People? I wrote about a payroll software company that lets 45 developers go, and how the restaurants it sells to end up with fewer guests. That was one company. If 90% of all office workers lose their jobs, the economy doesn't grow from all that productivity. It shrinks, and in the worst case it collapses.
Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.
Take a look at vroni.com