Conversation

Notices

1. Embed this notice
   Dana Fried (tess@mastodon.social)'s status on Monday, 17-Mar-2025 11:30:10 JST Dana Fried

   I don't know who I have to say this to, but:

   If you put your name on a professional product - whether that's a pull request, a legal brief, a business report, or anything else - you are responsible for the quality and correctness of that product.

   Corollary: I doesn't matter if you use an LLM, task an intern, hire a temp, or crib from Stack Overflow. You had better be able to vouch for the final product.

   If you cannot, you are making it clear that you cannot be trusted to properly do your job.

   • Embed this notice
     Dana Fried (tess@mastodon.social)'s status on Monday, 17-Mar-2025 11:30:15 JST Dana Fried

     in reply to

     If you're using an LLM to write code, you need to treat the output exactly the same as a random answer from Stack Overflow, or some shit an intern wrote.

     It might be great! It might miss corner cases! It might completely misunderstand the codebase and use APIs incorrectly, or just not compile, or solve the wrong problem!

     If you're employed as a software developer, it's *your* responsibility to proofread anything you didn't write yourself (and also, to be fair, the stuff you did actually write)

     Peter Krefting repeated this.
   • Embed this notice
     Dana Fried (tess@mastodon.social)'s status on Monday, 17-Mar-2025 11:30:16 JST Dana Fried

     in reply to

     • mcc

     I used to say that the most important skill I had as a software developer was google-fu, because 99% of the time someone else had already solved the problem I was trying to solve (sorry @mcc - you seem to be perennially in that other 1%).

     But I never blindly submitted Code I Got From The Internet without at least trying to understand (a) what it did, and (b) why it worked. Also: (c) writing tests to ensure it actually did work.

     Sometimes the code snippet from SO was actually kinda bullshit.

     Paul Cantrell repeated this.
   • Embed this notice
     Kagan MacTane (he/him) (kagan@wandering.shop)'s status on Monday, 17-Mar-2025 17:18:43 JST Kagan MacTane (he/him)

     in reply to

     @tess

     Student: But if I take all the time to read, review, and understand the code the LLM gives me, it won't save me any time at all versus just writing it myself!

     Master: *nods sagely* Indeed. It will not.

     And in that moment, the student was enlightened.

   • Embed this notice
     Nicole Parsons (npars01@mstdn.social)'s status on Wednesday, 19-Mar-2025 19:59:59 JST Nicole Parsons

     in reply to

     • Kagan MacTane (he/him)
     • David Mankins

     @lain_7 @kagan @tess

     Be careful about the "Cost of Convenience"

     1. AI is an energy hog. It's driving up electricity utility rates for everyone.

     2. AI is insecure. Anything created by an LLM can be monitored & be sold to a competitor. Your IP isn't yours anymore.

     3. AI robs junior programmers of growth and suppresses their wages.

     4. AI is biased. Your ability to incorporate accessibility features etc are reduced.

   • Embed this notice
     David Mankins (lain_7@tldr.nettime.org)'s status on Wednesday, 19-Mar-2025 20:00:00 JST David Mankins

     in reply to

     • Kagan MacTane (he/him)
     • Nicole Parsons

     @kagan @tess @Npars01

     It’s still easier and faster to kibbitz/interrogate/review existing code organization as churned out by the LLM than to organize and compose that code in the first place.

     GreenSkyOverMe (Monika) repeated this.