Carl Sandrock

@gordonwells There is still deniability for shorter replies. There are practical issues about either who manages the "clearinghouse" for this and what the retention policy would be or what the standard API should be and how to searching them all. If it works cheaters will go local.

How refreshing github.com/binjr/binjr

I've been practicing sight reading with an iPad app for 4 years and tracking my progress sandrock.co.za/carl/2025/05/f…

Some folks asked if old climate models were just continuing observed linear increases in temperatures. This misrepresents what physics-based climate models do (they aren't curve fitting), but also ignores that in ~1970 there was little observed warming:

Scientists have been publishing climate models since ~1970. A good way to evaluate their skill is to compare what they expected to happen in the years after the model was published to observed climate changes. It turns out most models were pretty spot-on:

@gordonwells Who is the "they" in this? The artists that appear on Song Exploder? If so, I think there is a (relatively) modern trend of people expecting mostly "authenticity" from artists, which almost necessitates writing personal experiences, rather than some kind of universality.

transformer-circuits dot pub by Anthropic is maybe the most gorgeous custom made research publishing platform I've ever seen it's so gorgeous and has so many neat features triple the salary of the dude who built this and fire the web team, ask (beg) him to spend 20% of his time on it

Yesterday I had this idea that you could figure out stitch counts by finding an FFT peak from a guage swatch picture. It works!

@akshaykagrawal @themylesfiles @DylanMadisetti @marimo_io We use jupyter notebooks extensively in our workflow, with jupytext to get clean diffs. One of the things I really appreciate is having the results of the analysis available in the repository. How do you save the results of the analysis in the Marimo workflow?

I've spent the past 3 years working with @themylesfiles and @DylanMadisetti to fix Python notebooks — version with Git, run as scripts, reuse as modules. Why @marimo_io stores notebooks as Python, not JSON: marimo.io/blog/python-no…

@gordonwells Madhur Jaffrey has given my family so much joy with the ultimate curry bible. Just made the duck vindaloo the other day.

I am finding my "journey" to do more front-end work incredibly frustrating and unrewarding. Today I figured out part of the reason is because of the mismatch between my expectations based on seeing features everywhere and wanting them and my ability.

@rovarma Have a look at jc github.com/kellyjonbrazil…

Don’t think I’ve ever written as much string parsing code as I have while working on Linux support. /proc files, cmdline tool output, etc, so much stringly typed stuff. Feels bad knowing you’re really just doing binary data -> string -> binary data conversions all over the place

Control theory is way more important to fixing ML than people realize. Control theory is just continuous reinforcement learning after all.

One thing that I have learned about myself in the last 5 years is that I love remote work and don't ever want to work in an open-plan office.

gm

built an ai that sees 5 moves ahead in any conversation and tells you the optimal thing to say

@gordonwells I will write 10× instead of 10x and pronounce it "10 times" instead of "10 ex". But language changes and eventually I will just be that weird dude that uses ancient conventions.

@gordonwells Then you're getting less use out of modern LLMs than I am. Using LLMs for "simple search" is harder to nail. Their unique selling point is not retrieval but in generation. Simpler search engines couldn't help me create a website or rewrite a recording to agenda points.

I hate it when people say "this technology allowed a person to fly a drone using only their thoughts". That focuses on the wrong thing. It's really "this technology allowed a person to fly a drone without moving any part of their body" or something similar. It's not telekenesis.

There's a big difference between knowing the name of something and knowing something.