This is crazy www.gsqi.com/marketing-bl...

> Google’s ‘Scaled content abuse’ spam policy does NOT specifically say that AI-translated content is spam. If you leverage AI to translate high-quality and helpful content, then it’s ok

arathunku.com/b/2025-11-10...

This is the way #jj-vcs

nah, just user error 😀

Maybe not so crisp after bsky compression...

With rustler, we should be even able to get native bindings? 🙃

+1 for English results. It still bad as much as google, for localized searches :(

Websites are using LLMs to automatically serve translated content

Lots of websites are doing it on their end, but I'm still expecting search engines to navigate that. We'll only get more AI slop, AI generated content, copy-cats, etc.

Right now the best way to avoid that is not to use Kagi, but Gemini/ChatGPT/Claude Deep Research for localized searches.

With this, I'm FOR debug or at least info level in logs, not warning.

When something fails, I want to see it ALL, not try to run them again with different log level.

- On main, failures get automatically retried few times, and flaky tests are reported if it was one-off

> people get mad when it wasn't introduced by their change

Is super real and very annoying for everyone.

We do this for all rspec and mix tests

I do this with bash

gist.github.com/arathunku/32...

`sprogress` captures full output and prints it on failure / C-c.

On CI, this would be layered approach:

- Any changed spec file runs 3-5 times (reduce probability of flakiness)
- Run all specs with something like sprogress

`green` revset is amazing idea. I'd love to stack my changes, and have CI part of the output both for filtering and inside log too!