News for hackers. Stuff that matters.

samesense

Re:Databricks drove down AI coding spend 70% on Friday August 07, 2026 @11:20PM (#49218622)
You have an unlimited budget, and you only spend $80/day? I’m up to $3k/week, and still expanding.
Re:Ask HN: What Are You Working On? (July 2026) on Sunday July 12, 2026 @07:50PM (#48886104)
A clone of popurls that learns what I like to read: https://perurls.onrender.com/
I get that grouping similarly colored icons might look better, but does it make it harder to find the app you want? Do you mainly pick apps by position? For quickly locating an app, wouldn’t it be better to place similar colors far apart, so you remember, for example, that the music app is the red icon in the lower right?
Re:Blog Feeds on Saturday October 04, 2025 @07:27PM (#45477667)
I had a very similar idea. I’m glad someone implemented it.

Show HN: Top Neovim/iTerm theme pairings (color compatibility scoring)

I made a list of Neovim/iTerm theme pairings that look cohesive when used together. Instead of picking by eye, I scored them by color palette similarity (nearest-neighbor matching) and balance across backgrounds, highlights, and syntax colors.Top 10 Pairings:• Everblush × nightly.nvim• Tomorrow Night × rusty• Rose Pine × Sakura.nvim• Challenger Deep × DarkScene.vim• Atom One Dark × onepro•...

( Read More... | no comments | on HN )

Show HN: Comparing iTerm2 and Neovim Theme Similarity

https://rpubs.com/samesense/theme_colors

I compared iTerm2 and Neovim color themes by extracting their palettes from theme files and computing a similarity score. Theme pairings are recommended for users who want a consistent look across their terminal and editor, since their palettes are perceptually similar according to CIELAB-based distance metrics.

( Read More... | no comments | 2 points | on HN )
Re:Beyond the Wrist: Debugging RSI on Wednesday May 14, 2025 @03:08PM (#43988142)
I’ve been controlling rsi with upper body mobility and strength training + typing breaks for years. This mind over matter approach sounds like bullshit.
Here's a python script to find the rss url on a science journal's website. It leverages smolagents and meta-llama/Llama-3.3-70B-Instruct. The journal’s html is pulled with a custom smolagent tool powered by playwright. Html parsing is handled by a CodeAgent given access to bs4.

I've tested with nature, mdpi, and sciencedirect so far.

I built it b/c I tired of manually scanning each journal's html for rss feeds, and I wanted to experiment with agents. It took a while to get the prompt right.

Suggestions welcome.

I use snakemake along with dvc. It detects changes in files, and reruns steps to produce downstream files as needed.
Is this a tool, plugin or philosophy?
producerpower doesn't work for me
Re:Build a salary with GitHub Sponsors on Saturday January 23, 2021 @09:53AM (#25882620)
Excellent. I’d pay to see guido’s python.
Re:How Many Die from Medical Mistakes in U.S. Hospitals? (2013) on Saturday September 24, 2016 @08:26AM (#12570672)
Yes. I've seen several examples of decision support since 2010. It's not easy to train models because the codes for procedures can be ambiguous. Furthermore, since hospital classifiers favor medical error recall over precision, docs can be swamped with so many warning messages that they tend to ignore them.

Take adverse drug interactions as an example. The training data for drug interactions mostly come from adults, so the resulting models do not apply in a pediatric setting. When the models are let loose in pediatric hospital, a high percentage of the drug interaction warnings are false positives, so these type of warnings tend to be ignored.

It seems that the trend is to use decision support with a lot of human oversight and investigation of the raw data to see if the model conclusions are correct.