Our dev commands sometimes output authentication URLs in the terminal that need to be clicked, which then opens a form to submit.

Suddenly blocking on this annoyed me, so now I have:

1. An iTerm2 "trigger" to automagically open the URLs: https://github.com/henrik/dotfiles/commit/e34a9df0a08453615fea89cd8e53dea7737d9936
- An Arc Boost (userscript) to autosubmit the form once opened.

So nice!

The script tracks opened unique-code URLs to only trigger once, as URLs in a terminal can otherwise trigger *repeatedly* (infinitely)… ask me how I know 😅

Excited to discover the IKEA app (at least here in the UK) has audio pronunciations of the names… but they are wrong? Then what is the point? 🫠 #SLOPP

Of the 3 I tried, 2 were wrong:

VINTERFINT should presumably be /fi:nt/ (”winter pretty”) not /fint/ (”winter dodge”).

GÖRSNYGG should surely be a hard G (Gothenburg dialect ”very attractive”) not a soft J (”makepretty” as a nonsense compound). Also, Gothenburg dialect mandatory for this one.

cc @robinwhittleton

As someone mixing languages freely in my Obsidian vault (Swedish/English), I just realised I can use aliases to be able to find some files by either language (my "Mått" file now has a "Measurements" alias).

It would be tidier to use English for all filenames, but it's a second brain… it's not always going to be tidy.

https://help.obsidian.md/aliases

I had Claude Code visualise our RabbitMQ topology; I also asked questions about RMQ in general since I don’t know much about it.

This makes me wonder how well Claude might work to generate codebase-specific educational material – on general topics but with specific (to you) examples.

Creepy/convenient:

Used Apple Pay in a physical store I’ve never bought from before (online or in-store), and the Shopify point of sale terminal pre-filled my email for a digital receipt.

I thought Apple Pay sent different identifiers to different merchants to prevent tracking, but I guess all of Shopify’s network counts as a single merchant in this case?

(cc @atpfm)

I have a stronger drive than most developers I’ve worked with to share solutions. I hate the idea of someone else running into the same problem and not being helped by me having already fixed it.

I just realised now that AI agents are potentially the extreme form of this, for code that makes its way into their training data. Though perhaps it’s outweighed by eating its own slop.

Our "devs look at logged exceptions and urgent requests" role is called "The Hatt" for Reasons.

I realised today that I want a TUI for monitoring The Hatt and then realised it could (should) be called "hattatui" 😍