Post

Conversation

Sayash Kapoor
@sayashk
After we invented the dynamo, it took us 40 years to electrify factories. In the process, we had to redesign the entire factory layout — electrifying existing factories didn't cut it. Software engineering will likewise need to undergo drastic changes to truly benefit from AI. Using AI to speed up existing tasks might seem helpful, but redesigning how we write code/maintain repositories will probably be far more impactful. But as we discuss in AI as Normal Technology, this process is time consuming, and we will need broader adoption to figure out what will genuinely help.
Screenshot from AI as Normal Technology
read image description
Quote
METR
@METR_Evals
We ran a randomized controlled trial to see how much AI coding tools speed up experienced open-source developers. The results surprised us: Developers thought they were 20% faster with AI tools, but they were actually 19% slower when they had access to AI than when they didn't.
Image
Last editedOpens edit history
13.9K
Views
David Watson 🥑
Post your reply

Joel Becker
@joel_bkr
agreed that this is directionally a win for this kind of economicsy take! i do think an important difference, though, is that AIs are seemingly increasing possibilities much faster than electricity, so the need to 'redesign factories' as it were could be obviated
206
Ben
@Ben3adi3
Strong agree. We're already seeing a bit of it in a way I'm not a huge fan of. Fullstack Javascript is becoming the go-to for most startups because of how easily AI can generate the base app and because it requires only one repo. We're seeing a decline in traditional backend
46
Chicken
@youdontsayhmm
Comparing factories with software is just silly. I can install a new ERP solution in minutes. Once AI becomes ubiquitous and really advanced ripping out the old and putting in the new will become a 24 hour or maybe one week job.
69