>asked AI to convert tests>25% failed>Ask it to keep going until they all work>Ask the tests pass so that means it's workingHow long before this shit comes back to bite them in the ass?
https://www.softwareseni.com/how-airbnb-compressed-years-of-technical-debt-into-weeks-using-ai-coding-assistants/Forgot the link
buy an ad, newfag.you need to learn how to not link.
>>106755528>NewfagFuck off cunt. I've been stuck on this shit hole for 20 God damn years. I'll link if I want to. I prefer that to screenshot threads with no link to the usually fake tweet.
>>106755425what tests? and what is technical debt (i feel dumb not knowing this)all lower case because typing with one hand
>give framework to LLM to update code base >it does it because H1B's don't know or care how to do it outside their local machine>it updates the code baseWOW>>106755627Technical debt is basically the cost it would be for an external or internal staff to focus on completing a project.Example: We run everything on legacy PHP/Mysql and want to move to modern PHP and so on. The project takes 10 people, with 100k/yr salary, 1 year to test/dev/prod roll out. Project costs 1m of technical debt to do.While that's a heavy oversimplification of it, that's basically what it is.
>>106755627What that other anon said but with an emphasis on the debt part. Usually you run a project and come to a fork in the road. >Pajeet my son >It is time to decide>will you create a robust solution using modern tools that require you update your php version>Or will you forgo the update in order to deliver faster>Gets hacked on the designated shitting streetThe tests are unit tests. Testing units of code to make sure that it does what you expect. >Unit test the isEven function>Tests a few even and odd numbers>Test null values>Test strings>Test that the right response is served or the service errors when it shouldDoing this over with AI is a dumb idea because the AI can and will just change the test to "make the test pass". And if they avoided it because it was gonna take a year and a half chances are they didn't check to make sure the replaced tests were meaningful and correct.glhf
>>106755746Could be a good thing if the tests look good to a human with tasteI used AI to convert old-style class-based React components to the new-style function -based components And everything worked fineCleaned up a couple longstanding bugs, too
>>106755425nah using ai to write tests seems like a good idea tests are retarded, at my previous job the time on a ticket was like 30% working on the bug/feature and 70% writing tests its such a waste then we have all those tests and would always spend a week manually testing overthinking before a release anyway. at my current job im the only developer for the mobile app so i dont write any tests, hardly ever have bugs, i do manual testing when doing the work why do i need to write unit tests ontop of that i hardly ever have tickets come back from qa
>>106755574Archive link probably would have been better, but thanks for including a link at all. Screenshot threads are worthless.
>>106755786The probability of AI screwing something up hard is how far it has to go from A->B and how long the output is, so converting from one style to another in the same language is obviously a strong usecase. If I write some novel shit in C++ and ask an AI to convert it to assembly I'm basically asking it to be a compiler, a task it will fail miserably at.
>>106755425does nobody realize this article is ai-generated slop? the new internet sucks. so. much.
>>106756241https://medium.com/airbnb-engineering/accelerating-large-scale-test-migration-with-llms-9565c208023bBecause the original writeup is here. Don't know OP chose to link to a content farm instead.
>>106755574>20 years of 4chan and still stuck doing worthless screencap threadsKill yourself. A zoomnigger would be preferable to you.
