Iván Sánchez OrtegaOct 24

I'm in a #github internal group for high-profile FOSS projects (due to @leaflet having a few kilo-stars), and the second most-wanted feature is "plz allow us to disable copilot reviews", with the most-wanted feature being "plz allow us to block issues/PRs made with copilot".

Meanwhile, there's a grand total of zero requests for "plz put copilot in more stuff".

This should be significative of the attitude of veteran coders towards #LLM creep.

Show threadEivind JahrenOct 25

@IvanSanchez @leaflet I can understand not wanting the AI review feature, but reviewing with an LLM initially isn't that bad. It does better at reading every word of a 1kloc than a human. So if the docs are inconsistent with the code then an LLM is better at finding it.

You might say that 1kloc is a too large diff for a review. I think using AI to review a diff is sub par, better give the entire file. Humans can also benefit from looking at the entire file, but will likely start to skim.

Show threadIván Sánchez Ortega
@Haskellelephant I'm afraid you might be confusing "better" with "faster and without getting tired" 🤔
Oct 25 at 9:13amFedilab
Show threadEivind JahrenOct 25
@IvanSanchez To some extent those are the same thing. If you have a trivial pr and I say I will spend 20 days thinking it through thoroughly then clearly that is completely inappropriate. However it does depend on what is at stake, on a car you spend more time checking that the brakes work correctly than on the windshield wipers. When you can do some quality assurance more efficiently then you can get more quality. Whether you use that increased efficiency to do less QA is on you.