What’s hilarious to me about products like FireHydrant integrating an LLM to give status summaries and bootstrap retrospectives is that it just floods things with flowery vagueness and weasel words.
This is the exact kind of stuff that would get immediately flagged at AWS and asked to be rewritten. Reading a sentence like “the incident impacted a small number of customers that the team swiftly leapt in to resolve” is a complete waste of time.
Every time I’d reach for xargs in a shell pipeline crunching a lot of inputs I’d always worry about there being too many arguments and blowing past the system exec limits, and manually try to limit invocations with a guesstimate of how many max args to use
Except I just learned that xargs can not only inspect the system and figure out how many args to use safely, it does it by default!
Easily one of my favorite utilities that went up in my list today
Writing a GitHub Action be like: why do something in 10 lines of bash when you could write 200 lines of JavaScript and pull in 30 package dependencies instead
But seriously though, I don’t understand why GitHub decided the only blessed way to get post-hooks is to write an action that must be executed by NodeJS. You can easily write a JS script that executes a shell script, so why not cut out the middle man entirely?
Sure GitHub publishes some npm packages which give you some cross platform APIs, but for a simple, Linux-only CI a shell script is more than plenty
Oh man, if I had realized that nix-darwin can give me fish autocompletions for #Nix (including picking flake outputs) I would have started using it years ago just for that alone
One aspect of #NixOS modules no one ever talks about: if you fetch and import modules written by someone else, you are effectively trusting them with root access to your machine