I performed this experiment a few weeks ago with a language I do not know (zig) and a domain I am bad at (game development) and walked away with a cross platform multiplayer physics simulation
I started with o1 in cursor, and told it to generate an architecture design document,
then with cursor's agent composer, I dictated and debugged almost exclusively through the chat interface using sonnet 3.5
ive been using cursor to start an initial solution (and by initial I mean it really works fully)
and then I manually refactor and read every single line.
but I can that in 1 or 2 years from now.. code quality will start to be worse and worse over time. and then you will have a bunch of highly senior engineers trying to understand what happening with their AI code 馃槄
Yes... I think this will be up to us as engineers to fight against it else we will just end up being replaced by others who are able to not have any fun
The way I see it, engineers still have the energy and motivation to build, whether that be through prompting or "hand crafted" code, and the commodization of software will not replace the need for people who want to build
Even if you can prompt away a service to handle some task for you, there will still exist many people who don't want to deal with, maintain, host, ideate on that service, and just want someone to build it for them
