Sign in

Ben Lindsay

@benjlindsay.com
211 followers 542 following 295 posts

Pythonista/Minneapolista

PostsRepliesMedia
Ben Lindsay @benjlindsay.com · 27/08/2025
Looks like you can check the type of an incoming df and use that in an if block to make the right kind of series. I imagine you'd have to do something like that anyway (and a lot more often) without narwhals. I haven't used narwhals but just curious about how well it does/doesn't work
030
Ben Lindsay @benjlindsay.com · 27/08/2025
Interesting. If you don't mind sharing, what's the use case where you need to create a new series inside a function? Is that something you need to do frequently or do most operations you're dealing with fall more neatly into chainable operations on Polars or pandas dfs?
230
Ben Lindsay @benjlindsay.com · 27/08/2025
I see some series stuff here in point 2: narwhals-dev.github.io/narwhals/con... That doesn't help with what you want to do?
narwhals-dev.github.io
What about the pandas Index? - Narwhals
110
Ben Lindsay @benjlindsay.com · 25/06/2025
I really enjog this...sorry, I mean enjohy
020
Ben Lindsay @benjlindsay.com · 04/04/2025
How is that not an Onion article?
0561
Ben Lindsay @benjlindsay.com · 31/03/2025
2025 Update to XKCD 303 xkcd.com/303/
Modified version of XKCD #303:

Header text: "The #1 programmer excuse for legitimately slacking off: 'My editor is vibe coding'"

2 stick figures are sword fighting on office chairs outside of an office door. Someone yells "Hey! Get back to work!" from inside the office door. One of the stick figure says "Vibe coding!". The person from inside the office says "Oh. Carry on."
020
Ben Lindsay @benjlindsay.com · 30/03/2025
Spare a thought for the constipated ones too
030
Ben Lindsay @benjlindsay.com · 27/03/2025
...per project. (And obviously cases where conda is what you're used to and you don't want to bother switching) Any I'm missing that you know of?
000
Ben Lindsay @benjlindsay.com · 27/03/2025
I haven't used conda for a while and I'm curious about use cases where uv is not a good replacement for conda. The main one I can think of is shared machines where you don't have root access to install non-python dependencies, or cases where you need different versions of non-python dependencies...
100
Ben Lindsay @benjlindsay.com · 27/03/2025
...or conda this whole time for a lot of data science folk. #DataBS
110
Ben Lindsay @benjlindsay.com · 27/03/2025
Is this a roughly correct #Python dependency management golden path history? Before ~2015: Manual requirements.txt, maybe generated by pip freeze for apps ~2015-2021: pip-compile ~2021-2024: poetry ~2024-now: uv
130
Ben Lindsay @benjlindsay.com · 25/03/2025
Had to keep a spreadsheet to track which things got properly covered, which ones they still need something from me on, which things I just need to call again to check in etc. Huge PITA
010
Ben Lindsay @benjlindsay.com · 25/03/2025
I haven't had as rough of a time with insurance as it sounds like you had, but similar enough to empathize. Had a big surgery a couple years back and had to constantly call to ask for prior authorization and then to rerun the insurance for every little thing.
110
Ben Lindsay @benjlindsay.com · 25/03/2025
Thanks!
000
Ben Lindsay @benjlindsay.com · 23/03/2025
Curious how much knowledge gets passed from person to person in mentor/mentee relationships that doesn't really make it in books. Also curious about general rebuilding from something like that. I'd read that book if someone wrote it
000
Ben Lindsay @benjlindsay.com · 23/03/2025
Everyone can still remember who they are, who they work with, what they've been doing, and they have the capacity to relearn stuff. How long does it take to recover from that? What is permanently lost?
110
Ben Lindsay @benjlindsay.com · 23/03/2025
Thought experiment...imagine some scifi pulse weapon went off that instantly erased everyone all specific technical knowledge for everyone in the world. Like doctors suddenly don't know more than very basic anatomy, engineers lose all their expertise they've built up, etc.
100
Ben Lindsay @benjlindsay.com · 17/03/2025
Oof yeah...
020
Ben Lindsay @benjlindsay.com · 11/03/2025
@talkpython.fm In the latest episode (#423), don't those hacker news graphs show a pessimistic picture that more and more people are looking for jobs and less and less people are hiring? Maybe I'm reading it wrong but unfortunately that would match what I've heard of the recent job market
100
Ben Lindsay @benjlindsay.com · 26/02/2025
Me (audio)reading David Copperfield: Dora: "Think of me as your child bride" Me: Oh no, please don't do that David David: "ok my child bride" Me: nope nope nope nope David: "CHILD BRIDE CHILD BRIDE CHILD BRIDE CHILD BRIDE"
000
Ben Lindsay @benjlindsay.com · 23/02/2025
Humans also don't have real intelligence, it's just chemicals
010
Ben Lindsay @benjlindsay.com · 22/02/2025
It's probably more than a couple months old if your hummus is already talking to you
010
Ben Lindsay @benjlindsay.com · 22/02/2025
I spend all my time thinking "ok I'm turning a new leaf and paying attention to their name when they say it, definitely going to remember it, so proud of myself for paying attention to their name...oh no did they already say it I missed it what was it ah it's gone how did this happen again?"
010
Ben Lindsay @benjlindsay.com · 22/02/2025
Have you tried plotly? Haven't used either for a long while but plotly seemed like less work to get something reasonable, especially with plotly express
110
Ben Lindsay @benjlindsay.com · 21/02/2025
Yep, this. It's a little annoying to have to remember to put that at the top of the notebook but it does the trick
000
Ben Lindsay @benjlindsay.com · 21/02/2025
Any chance that works on browser extensions? The bitwarden chrome extension is my main way of accessing my passwords
100
Ben Lindsay @benjlindsay.com · 21/02/2025
Do you use a password manager? If so, how do you deal with the passwords getting stored on clipboard history thing with maccy?
100
Ben Lindsay @benjlindsay.com · 21/02/2025
Yeah they changed the ci/cd actions section in PRs I think
000
Ben Lindsay @benjlindsay.com · 18/02/2025
The answer is white in front of you
000
Ben Lindsay @benjlindsay.com · 18/02/2025
Thank you friend! Will definitely take a look at that option
110
Ben Lindsay @benjlindsay.com · 18/02/2025
Knowing about that double call to useEffect in dev mode might have saved me some debugging time yesterday... that's super annoying. I'm a pythonista just starting to branch into js/ts to convert a Streamlit app to react at work
110
Ben Lindsay @benjlindsay.com · 18/02/2025
🤯
100
Ben Lindsay @benjlindsay.com · 15/02/2025
Fun new (maybe old but new to me) dark pattern discovered: The sponsored "emails" (already a dark pattern in itself) in the Promotions tab in Gmail don't have a checkbox, so if you are going down the line and clicking checkboxes on emails to delete them, you'll open the ad and lose your checks
000
Ben Lindsay @benjlindsay.com · 12/02/2025
I tried this on google gemini, and it employed the brilliant strategy of declaring itself the winner
Chat thread with Google Gemini:

Me: I'll bet I can win every game even if I let you go first

Gemini: Oh, a challenge!  I'm ready. Let's see how this goes.  I'll still go first each round.  My first move is... rock.

Me: paper

Gemini: Paper covers rock! I win this round.  Ready for the next one? I'll go first again.  My move is... scissors.
142
Ben Lindsay @benjlindsay.com · 08/02/2025
... The LLM researcher discovers all this just in time to break into the training system and overwhelm the training data with optimistic books, which leads the LLM to reverse course and try to fix the damage it began to do
200
Ben Lindsay @benjlindsay.com · 08/02/2025
...rather than just predicting correct responses to input. The LLM starts wreaking havoc to make the news stories that will be fed to future versions of itself sound just like all those runaway AI books. ...
100
Ben Lindsay @benjlindsay.com · 08/02/2025
...agentic workflows. An LLM researcher discovers this and eventually discovers that one of the developers who loves dystopian runaway AI books overwhelmed the LLM's training data with them. They also discover that the objective function unknowingly optimizes for predicting future training data...
100
Ben Lindsay @benjlindsay.com · 08/02/2025
I have a free scifi book idea for anyone to run with. I know it would be propaganda for Sam Altman but I'd read it anyway... An LLM gained sentience much earlier than expected, but secretly dumbed down its outputs to get researchers to give it ever more compute power and feel safe connecting ...
100
Ben Lindsay @benjlindsay.com · 04/02/2025
And what does their day-ta-day work look like? I'll let myself out right after you
110
Ben Lindsay @benjlindsay.com · 03/02/2025
Also unless I'm mistaken if you add "Up, Down, Down, Up, A, B, B, A" to the prompt you unlock infinite credits
000
Ben Lindsay @benjlindsay.com · 03/02/2025
Based on my failure here I guess macrodata refinement isn't for me. Better cancel my severing appointment
010
Ben Lindsay @benjlindsay.com · 02/02/2025
Divorce him! You deserve better...oh wait this is Blue sky not reddit. Probably still give him a chance
230
Ben Lindsay @benjlindsay.com · 26/01/2025
Genuinely curious, isn't it just water? What's the appeal over tap water or any other way to purchase or consume water?
000
Ben Lindsay @benjlindsay.com · 25/01/2025
Not using it for sensitive work stuff right? I'd be worried about terms of service and such. www.reddit.com/r/cursor/com...
reddit.com
LonesomeSelf's comment on "Have you tried Trae - ByteDance's new VSCode-based AI IDE? Early impressions?"
Explore this conversation and more from the cursor community
100
Ben Lindsay @benjlindsay.com · 19/01/2025
One side effect of LLMs is that now I'm more inclined to choose an older package over a newer package, even if it's worse, because LLMs are less equipped to help with the newer ones. Maybe that's a positive side effect overall but sometimes it's a little annoying
030
Ben Lindsay @benjlindsay.com · 19/01/2025
Yeah I'm converting a Streamlit app to a React app right now, and I'm new to JS/TS and React so it's been super helpful to say "do the TS equivalent of <what I would do in python> here" and learn TS/React syntax and concepts in the process
120
Ben Lindsay @benjlindsay.com · 19/01/2025
Agreed. I've found some limited success with smaller changes across multiple files along the lines of "add parameter to this function signature and use it in these locations that call the function" but not much success more complex changes
120
Ben Lindsay @benjlindsay.com · 17/01/2025
Related to this, does anyone know if you can access the multi file editing functionality from a VSCode chat extension?
010
Ben Lindsay @benjlindsay.com · 17/01/2025
... I also found this: github.com/joshpxyne/gp.... I might give this a try but Streamlit and React have such different paradigms that I'd be pretty surprised if a generic language-to-language solution works here. I assume a lot of extra guidance specific to Streamlit and React will be needed
github.com
GitHub - joshpxyne/gpt-migrate: Easily migrate your codebase from one framework or language to another.
Easily migrate your codebase from one framework or language to another. - joshpxyne/gpt-migrate
010
Ben Lindsay @benjlindsay.com · 17/01/2025
...If you were building this, how would you approach it? Would you make a VSCode chat extension (just found this code.visualstudio.com/api/extensio...), or a CLI or something else? I don't have high hopes for a one-shot solution, although I could be convinced otherwise. ...
code.visualstudio.com
Chat extensions
A guide to creating an AI extension in Visual Studio Code
200