Notices by Ellie Huxtable (ellie@hachyderm.io)

A friend of mine made a really cool thing to detect blocked tokio runtime threads

You should check it out: https://github.com/conradludgate/tokio-supervisor

Another sneak peek of what I'm working on! Can't wait to share a build :D

"Always code as if the guy who ends up maintaining your code will be a violent psychopath who knows where you live"

- overused
- negative
- unnecessarily violent

"Always code as if your code will be showcased in a prominent blog"

- fresh perspective
- motivating instead of terrifying
- pleasant and wholesome

A fantastic post from Julia Evans here! https://jvns.ca/blog/2024/07/08/readline/

I never thought that giant ugly match thing Atuin has would be linked to from a famous blog 😂

I now block several known LLM crawlers from my site

https://ellie.wtf/notes/blocking-ai-crawlers

If you'd like to do the same, feel free to take my robots.txt

@mo8it Your PR example makes total sense, but when it's on something like a feature request it's especially annoying

We end up getting a whole bunch of them every single week, and something that's important to that _one_ person may not be important to the pool of users as a whole.

Also it's just not fun to get a reminder than your software has problems/shortcomings for someone while you're sat watching TV or out for a walk or something. I don't need the ping for that

If you ever find yourself writing an issue comment like

> any news?

> plz add

> this is a dealbreaker

Please... don't. Give the op a 👍, and move on. It'll show up when maintainers sort by reactions, and avoids mildly unpleasant emails in our inboxes.

Serverless replaced servers ✅
Crypto replaced money ✅
LLMs replaced developers ✅

Will tech go to the gym for me next?

there's three options I see

1. No telemetry. Hope that people report things accurately
2. Opt-in telemetry. Hope that enough people opt in to give something statistically accurate
3. Opt-out telemetry. Gives the most data, but can be pretty creepy

Issues with (1) and (2) are that only those who are positively engaged + more technical are likely to opt in or report.

Issues with (3) have been widely discussed all over already

touchy subject! but does anyone have any examples of telemetry in open source projects that has been done well?

success criteria:

1. allow the project to allocate time to things people actually use
2. identify issues people may have, as not everyone files issues or can effectively report a bug
3. preserves privacy + anonymity, and isn't creepy

eg right now I'd love to know what config/features people actually use, so I can stop maintaining what isn't necessary

@linus I wouldn’t go that far. Especially for juniors, all they hear is that AI is the next best thing and enables them to work better and faster.

It’s understandable that they may lack the judgement and experience to see that that’s not quite correct

I’d love it if AI made my life easier, but so far

1. Package managers that are straight up misrepresenting my work
2. Chat bots advising people how to use my software incorrectly
3. PRs that just don’t quite feel right

zsh fixed, it's fast again

https://ellie.wtf/notes/profiling-zsh

It saddens me how much knowledge is locked away in Discord servers, and will absolutely vanish within 5yrs

Forums from 20yrs ago still show up as useful for me from time to time

What’s your worst ops blunder?

Mine is the time I took down an entire AWS account by miss-spelling an environment variable 😳

Orrrr the time I accidentally deleted all the code from self hosted GitLab right before end of day

#infrastructure #devops

My neighbours have claimed that my van “spoils their view” while parked on the street with everyone else

Any suggestions for how I can make the view better?

I was thinking of wrapping it bright purple but I’m not sure that would be enough of an improvement

When I make a mistake, I'm "wrong"

When GPT-4 does, it's "hallucinating"

I'm not wrong, just had something extra in my tea this morning ?♀️