Some observations.

Some observations. In my feel, automating coding will lead to an interesting effect:
1/ the engineer profession will get closer to the scientist profession — meaning where ideas and understanding are pricier than applied skills (that’s actually how it was way back in the 70s–80s)
2/ the amount of trash software will be huge, but nobody will need it. This is that thing where every product person or mediocre engineer rushes to build “products.” Yeah, there’ll be more of them. But there’ll be a TikTok effect: when the choice is that huge, people will pick the best. And the best requires a higher level of approach.
3/ overall nothing will change, I think. Yeah the feedback loop will get faster, yeah iterations will get cheaper, but that’s all on top of “so what the hell are we even supposed to do”

Way more interesting is to watch whether progress shifts in theoretical digging (research automation). Because so far it’s not that cool there. But there are movements.

Cursor is building its own browser from scratch with agents
Dario from Anthropic, as usual, predicts the death of regular engineers
Write code better than Claude — a Challenge from the Anthropic team

Links from the original post:

• https://cursor.com/blog/scaling-agents
• https://www.businessinsider.com/google-deepmind-anthropic-ceos-ai-junior-roles-hiring-davos-2026-1
• https://github.com/anthropics/original_performance_takehome