Your friendly neighbourhood RSE

@tinashemtapera.bsky.social

Research software engineer #RSE @Harvard #rstats | #python | climate science | public health Research Software Engineering for Reproducible Science

“Though LLMs originate from a hyperactive tech sector + late stage capitalism, their domain of innovation may better fit knowledge and information access, than yet another app subscription. ITT, we propose that your local librarian should be given full&complete control over all LLM access and use…”

While I personally don’t see the appeal for Typst since everything I’ve ever done is in R and Python, I’m going to evangelise computational notebooks regardless of the language. More software — especially for data science of all flavours — should be created in notebooks. It just makes sense

Vincent Arel-Bundock@vincentab.bsky.social · 2mo ago

Yep! Quarto can do many many things, obviously. But if you'd like to mix code and prose in a notebook or website, and if you'd like to write in Typst (which is very, very nice!), then Calepin might be a good alternative for you.

At this point I think I wanna become an expert in software engineering in academia and HPC just to prove it can be done… it’s so bloody irritating how data scientists in industry are like, “just develop in a container” and “just get an AWS account” and all their problems go away

I’m more and more convinced every day that I work in technology, especially with code, that you should be using instant messaging (like slack) to do debugging and troubleshooting sessions.. it just doesn’t make sense to zoom code and not have any of your commands recorded

Istg this kinda thing is beyond frustrating not because it’s not a burden to create reproducible science — it is, but what’s more irritating to me is that this burden actually gets lighter the earlier you implement your reproducibility strategy…

Paul Pharoah@paulpharoah.bsky.social · 3mo ago

#statsky #episky I'm reviewing a paper and a link is provided to a github with the analysis code. There are several R scripts (none annotated) and the main, complex analyses clearly not included in any of them. The data are available "on reasonable request". Is this a red flag? Does it matter?

In today’s hypercapitalist Internet landscape every single thing that you experience — voluntarily or not — is to turn you into a customer… That’s why I’m happy to be working with today’s sponsor, outside! I’m not selling anything, I just wanna remind you to go outside

Kevin Reuning@reuning.bsky.social · 4mo ago

Just got a podcast ad for cornhole, as in the game. Not a specific place to play, just that you can play it.

I have to update one of the most inflammatory things I said back on the bird site, but for the LLM era: 🔥 🔥 🔥 🔥 If your agent prefaces every tool call with `cd /the/current/workspace &&`, I will come set your data center on fire 🔥 🔥 🔥 🔥