←back to thread

Some things to expect in 2025

(lwn.net)
258 points signa11 | 1 comments | 16 Jan 25 22:55 UTC | HN request time: 0s | source
Show context
kirubakaran ◴[17 Jan 25 00:45 UTC] No.42732804[source]▶
> A major project will discover that it has merged a lot of AI-generated code

My friend works at a well-known tech company in San Francisco. He was reviewing his junior team member's pull request. When asked what a chunk of code did, the team member matter-of-factly replied "I don't know, chatgpt wrote that"

replies(16): >>42733064 #>>42733126 #>>42733357 #>>42733510 #>>42733737 #>>42733790 #>>42734461 #>>42734543 #>>42735030 #>>42735130 #>>42735456 #>>42735525 #>>42735773 #>>42736703 #>>42736792 #>>42737483 #
deadbabe ◴[17 Jan 25 01:38 UTC] No.42733126[source]▶
I hope that junior engineer was reprimanded or even put on a PIP instead of just having the reviewer say lgtm and approve the request.
replies(2): >>42733168 #>>42733515 #
WaxProlix ◴[17 Jan 25 01:44 UTC] No.42733168[source]▶
Probably depends a lot on the team culture. Depending on what part of the product lifecycle you're on (proving a concept, rushing to market, scaling for the next million TPS, moving into new verticals,...) and where the team currently is, it makes a lot of sense to generate more of the codebase by AI. Write some decent tests, commit, move on.

I wish my reports would use more AI tools for parts of our codebase that don't need a high bar of scrutiny, boilerplate at enterprise scale is a major source of friction and - tbh - burnout.

replies(3): >>42733431 #>>42733542 #>>42736479 #
not2b ◴[17 Jan 25 02:42 UTC] No.42733542[source]▶
Unless the plan is to quickly produce a prototype that will be mostly thrown away, any code that gets into the product is going to generate far more work maintaining it over the lifetime of a product than the cost to code it in the first place.

As a reviewer I'd push back, and say that I'll only be able to approve the review when the junior programmer can explain what it does and why it's correct. I wouldn't reject it solely because chatgpt made it, but if the checkin causes breakage it normally gets assigned back to the person who checked it in, and if that person has no clue we have a problem.

replies(2): >>42735128 #>>42737983 #
1. solatic ◴[17 Jan 25 14:42 UTC] No.42737983{3}[source]▶
Not being willing to throw out bad/unused features is a different trap that organizations fall into. The amount of work that goes into, shall we say fortifying the foundations of a particular feature, ideally should be proportional to how much revenue that feature is responsible for. Test code also has to be maintained, and increasing the maintenance burden on something that has its own maintenance burden when customers don't even like it is shortsighted at the very least.