"Deskilling" Shock is Coming | Anthropic Economic Report

I write a lot of my legal briefs with AI now. The quality improvements over the past year are amazing to watch and it boosted my productivity by several magnitudes as constructing/decunstructing arguments and legal analysis of Court decisions are now well handled by AI. This used to be a very time consuming part which got reduced by a lot. I still need to proof-read the results, do some corrections here and there. And dealing with a lot of lengthy texts can be a challenge for the context window with a sharp drop in quality past 100.000 tokens. But what used to take several days of work got reduced into a couple of hours.

Y’all are about to experience 40 years of automation of blue collar jobs compressed into 5 years of automation for white collar work. As a machinist who’s worked in industry since before widespread use of computers in factories, I’ve experienced the gradual replacement, offshoring, deskilling, and productivity leaps over my career. I easily do 5x more work and do it better than when I started. Less skilled workers do better work than highly skilled workers 40 years ago. But this was a long slog. It’s happening so much faster because the physical infrastructure of hard goods manufacturing is vast, expensive, and durable. You can’t scrap a $400 million dollar production line to make it 10% more productive. But with AI and digital work, that sunk cost is negligible. Reducing head count has always been the number one way to reduce cost. Even factories in the past would rather pay me time and a half and production bonuses rather than hiring more workers. It was cheaper.

i watch videos like yours while agent is doing my task in a minimized window

We spent all of 2025 saying that post-training would work amazingly well for programming and math, for a bunch of reasons. We’re now forgetting those reasons and acting as if it’s going to be just as “easy” for everything else.

Hallucinagentic Automuhtation

Wes “Explain it to me like I’m 5” Ros ;)

People try to creatively advertise AI as general tool for everything and keep failing, so there is good a reason for AI stagnation in Anthropic's report. There are so many great use-cases for narrow professional AI, that you can train for your needs.

I am 75 years old. When I was in college, an old man imparted a bit of wisdom to me. If I wanted to get the right answer, I had to ask the right question. Wow, how that applies so much for today.

10:48 That moment where you realize Wes is way way smarter than he usually sounds!

Your video quality and titling seems to have increased lately! I'm enjoying it more. Thanks for the work you put in.