The support the Python community has shown to PSF for its decision has given us hope in the sustainability of open science communities. Please consider extending your generosity to our organisation’s fundraising efforts, and help us to reach others aligned to our mission: carpentries.org/support/

"Crucially, students will need to develop sufficient expertise to identify when AI systems produce plausible but incorrect outputs — which poses a dilemma because this requires the very skills that AI is starting to replace."

www.nature.com/articles/d41...

Day 5: Maybe #LyricsPrompt #LyricsPrompt2025

The Gaslight Anthem: High Lonesome
www.youtube.com/watch?v=__ki...

"And at night I wake up with the sheets soaking wet" /
It's a pretty good song /
Maybe you know the rest?

I'd also like to shout out @carpentries.carpentries.org for leading the way on this, having made a similar decision earlier this year.

Thank you, Joelle 🙌

Put another way: in my experience it is still rare to find people sharing stories that are not (roughly) either "I tried this and it worked!" or "I tried this and it was useless!"

Interesting read, thanks. It chimes with my own experience and with other examples that I have seen people share. I was particularly interested in your case because you described a situation where the result of your first attempt was messy and needed a lot of manual work to fix.

That's my interest as well. Currently supporting Carpentries community members in the creation of a workshop teaching careful and effective use of genAI for (research) software development, and how to think about the risks/implications of that use. Case studies like yours are very informative.

Super interesting, thanks.

Right. Thanks for the detailed response. Since you mention that the model "guessed" what the input data would look like, could more context also have been helpful? E.g. what input will be and what output should be expected?

Reflecting on this experience, what will you do next time? Take the same approach, on the grounds that you have improved your debugging skills now & you estimate that this still saved you time? Or is there something you would adjust about the way you prompted Claude? Would you avoid altogether?

Finally, I should cite some related papers
1. From @baym.lol , a beautiful paper looking at genomes going back to the 1880s, showing how chromosomes gained AMR, with a very nice analysis of how AMR genes were mobilised and embedded in MGEs
www.microbiologyresearch.org/content/jour...
19/n

Great thread, thanks for posting it. And congratulations on the paper.

Error code: √-1