3 ms·

Obviously AI isn’t wholesale replacing anyone, but having a tool that can coach me through some library I’ve never used before can be the difference between a l

Obviously AI isn’t wholesale replacing anyone, but having a tool that can coach me through some library I’ve never used before can be the difference between a lot of trial and error and a 2 hour PR to finish a React feature that no one else is going to do. Whether that work (web dev, broadly) has real value or not is another question

n8cpdx 1y ago
Maybe, but that might fall into the bucket of "doing things that would have gone undone". AI slop often doesn't displace designers because it is used in places you would never buy a design for. If the react feature wouldn't be done without AI... It just wouldn't be done and the business would probably survive. Every software project has an infinite list of feature requests, bug fixes, etc. The cost of doing the work can shift a lot before it makes sense to start firing people. If there were actually a 100x productivity improvement it might be noticeable, the realistic estimate is closer to 20-30%, maybe 50 if you're writing very simple React exclusively. I think staffing levels is driven more by business/HR constraints than some ticket/worker calculation.
fhd2 1y ago
What I find curious in these AI discussions is how the tables have turned: Better tooling used to be pretty universally appreciated. I've never heard anyone say that IDEs, garbage collection, static code analysis, profilers, debuggers, frameworks or anything else that makes the life of a developer easier was "replacing" them. This stuff replaces programmers just like LLMs do, they make the job easier/faster for experts, so you can do more with less. They also make the job easier for laypeople, but there's specialised tooling for that already, mostly in the form of no code tools. Personally, I'd take an AI generated code base a non-developer whipped up over the no code nightmares I've seen _any_ day. But just like better tooling for experts, better tooling for laypeople isn't a particularly new thing.

n8cpdx 1y ago

Maybe, but that might fall into the bucket of "doing things that would have gone undone". AI slop often doesn't displace designers because it is used in places you would never buy a design for. If the react feature wouldn't be done without AI... It just wouldn't be done and the business would probably survive. Every software project has an infinite list of feature requests, bug fixes, etc. The cost of doing the work can shift a lot before it makes sense to start firing people. If there were actually a 100x productivity improvement it might be noticeable, the realistic estimate is closer to 20-30%, maybe 50 if you're writing very simple React exclusively. I think staffing levels is driven more by business/HR constraints than some ticket/worker calculation.

fhd2 1y ago

What I find curious in these AI discussions is how the tables have turned: Better tooling used to be pretty universally appreciated. I've never heard anyone say that IDEs, garbage collection, static code analysis, profilers, debuggers, frameworks or anything else that makes the life of a developer easier was "replacing" them. This stuff replaces programmers just like LLMs do, they make the job easier/faster for experts, so you can do more with less. They also make the job easier for laypeople, but there's specialised tooling for that already, mostly in the form of no code tools. Personally, I'd take an AI generated code base a non-developer whipped up over the no code nightmares I've seen _any_ day. But just like better tooling for experts, better tooling for laypeople isn't a particularly new thing.