My brain refuses to read LLM-generated text. I'm seeing agent-written snippets (particularly PR comments) that are just absolute word soup. I demand proof of life.

Mastodon tip that I saw someone post recently: you can use the search term "in:library" to specifically search among your own posts, likes and bookmarks.

The method has a name — Reservoir Sampling! (Thanks @shapr)

https://en.wikipedia.org/wiki/Reservoir_sampling

^ Days since I saw a singly-linked list indexed by rand(length)-1: 0

Programmers! Did you know you can fairly choose a random item from a list in a single pass, without knowing the total number of items in advance, and in constant, minimal space? This delightful method has lived rent-free in my head since I learned it from the Perl gods in the 90s: https://csg.sph.umich.edu/chen/Perl/cookbook/ch08_07.htm

Just received this wonderful #needlepoint gift from my daughter. Need to frame it for my office wall.

I've got a sunny outdoor space that is 61.8% as large as the house. It's the golden patio.

One of these years I'll actually make it to #fosdem...

The root of so many problems in software: reflexively upscaling capacity instead of addressing efficiency. Once you see it, it's everywhere.

- Process running out of memory? Don't just give it more, find out why first, and reduce how much it allocates.

- Team delivering too slowly? Don't just expand the team, find out what's slowing them down.

- Too much work to track? Don't just buy Jira and curate a huge backlog, identify the essentials instead, and keep a simple list

It's extremely magical thinking to assume that using someone else's reliable managed service or platform will automatically make your system more reliable. If it makes your system more complex, it probably also makes it less reliable, though the failure modes might be less obvious.

There's apparently a race to build the dumbest timeline on the dumbest timeline.