I love the work Alexander has been doing. His substack is a must read. Every single post is super well developed and starts important conversations.

Here is where we disagree on this series of posts: both academic articles and the work of juniors/RAs are equilibrium objects. Yes, AI can (or soon will be) able to write the types of articles we've written in the past. That means that new articles will just be better. Many of those working heavily with AI have seen their workflow, idea generation, research process change drastically with these tools. They are writing very different papers. The alpha will be in writing papers with AI that AI cannot do on its own. From my own experience: I am not writing more papers, I'm writing very different papers, ones I would have not been able to write before. This is incredibly exciting (for me, maybe not "the world").

Regarding juniors and RAs, I have to admit there was a few weeks in January where I had little work for my RAs. I did think: maybe there will no longer be an RA role. But then I sat them down, taught them how to use Claude Code, then Codex, gave them access to it, and now they're busier than ever (Jevons paradox for work). So yes, the types of things they were doing before have been largely replaced by AI, but RA + AI is incredibly useful.

All of this to say that science will change. I'm very worried about implications of AI for labor market, but not for science: it will be a huge positive and boon for humanity. (all of my statements above are not future proof btw, the centaur phase may be more temporary than I expect).