/TECH

Developers Rewrite AI-Generated Code Due to Overly Complex Abstractions

Commentary on X

Highest ranked

goes into the same direction as this tweet: models are good at producing working + (overly) safe code, which is incomprehensible and explained badly + wrong to humans. don’t think it’s a good direction https://twitter.com/kentonvarda/status/2074924213983740233

Combined views

8.6K

2 posts, first seen 7h ago

89 likes4 comments

Peter@peter58587TECH

@xeophon I find the tests they write are extremely poor. For every 10 tests, you might get 2 or 3 (maybe) useful ones. They're also extremely code additive. They're unable to write terse / beautiful code - not even Fable improved on this.

Florian Brand@xeophonTECH

working on a big rewrite and i can only agree current gen of models are very capable but also writing abstractions that are too sophisticated and complicated to understand (and also often not needed). i often find myself discarding the code these days and then re-doing it https://twitter.com/damian_b/status/2075019993402761597

6 saves

5 reposts

Sentiment

Positive0%100%Negative

Positive Read

Many users criticize AI-generated code for producing overly complex abstractions and poor tests that force developers to rewrite it.

Based on 2 visible X reactions from 3 accounts; directional sample.

Digg Deeper

Ask a question below.

Published answers will appear here.