/show-me is a phenomenal skill
Makes PR descriptions extremely easy to read
Basically a toolbox of "nice ways to look at code"
Nice work, @dexhorthy
https://t.co/phB1dAPwgG
@vanilagy@LewisCTech > The more that happens on the client, the better the user experience
progressive enhancement is also good (if you do it properly)
Absolutely love the /show-me skill by @dexhorthy for working on UI changes
1. Ask for an isolated HTML prototype of the thing you want to change
2. Iterate on it until you’re happy
3. Summarize the changes and apply them to the real UI
You keep the original reference in sight while exploring, without messing with the actual implementation too early
Please do not direct your agent to burn down issues on someone else's open source project. Reviewing your "contribution" is actually more work for a maintainer than it is for them to direct their own agent to fix the same bug. So you are actually creating work, not helping.
I remember 13-year-old me installing Ubuntu after using Windows 8. It was a breath of fresh air. Happy to see @dhh pushing Linux on the desktop.
But there are still some workflows Linux just can’t replace for me.
I like to make music in Ableton Live. I like to code. Today, macOS is basically the only OS where I can have both without compromise.
Will Linux desktop audio get good enough that Ableton, plugin makers, audio interface vendors, etc. start taking it seriously?
I doubt it. But I’d love to be wrong.
Tok/sec is a vanity metric.
GPT sol-low completes @ArtificialAnlys tasks in 36 sec/task
Qwen 3.8 27B at xhigh (same intelligence score( needs 13.2 min/task (+22x) w/+15x as many output tokens.
Qwen at 500 tok/sec (+10x) will STILL be slower and will cost +3-5x.
To show how much I care about performance, I just ordered the worst specced 2020 Intel MacBook Air I could find (i3, 1.1 GHz, 8 GB RAM). It is the last supported hardware for macOS 15, our min target for Superlogical. Going to optimize on it for worst-case behavior.