Posts from May 25, 2026

Something has changed with Antigravity since I first tried it out the other day. While looking at adding a related posts feature to BlogMore, I thought I'd give it another try out (having gone back to using Gemini CLI while I still could).

That first (and last) time I tried it, while on whatever model it decided for me out of the box, it chewed through most of the quota, with a 5 hour reset, in very little time at all. It was obvious that I'd never get anything of significance done in a good session.

This evening has been quite different. It wrote a very comprehensive change, doing quite a lot of work, and left me with a lot of quota and a short reset time once done.

A bit more testing and tweaking of the documentation followed, with me setting it off on a couple of bug hunts (which it found and fixed). By the time I was happy to call it an evening on this round of modifications, it had reset and I was green across the board again.

Now this I can work with!

I don't really know what's changed¹, or why. I think I saw something the other day about quotas being tripled, but this seems even more generous (at least in terms of the reset window). I guess I'm going to have to go digging to see if I can find what the story is.

I'm not getting my hopes up -- what can be given can be taken away at any moment (which is, of course, the ongoing theme of what I'm documenting here) -- but this does soften the landing somewhat.

I wasn't made for this. I didn't move to Scotland for this. I don't like this.

Yes, I know it's warmer elsewhere. Yes, I know it's warmer elsewhere in the UK, today.

Whatever.

This is still not fun and I still don't like it.

A quick update to BlogMore that stems from a couple of accessibility concerns raised when using PageSpeed Insights. While the accessibility score on pages generated by BlogMore is pretty excellent (96/100), a couple of things were highlighted that seemed worth tackling in some way.

The first was that the title bar of code blocks is rather low in contrast, especially the name of the language. This was a deliberate choice on my part; when I'd prompted Copilot to do this work some time back I'd expressly asked it to make the text "subtle". While I could have undone that, I do like how it looks and would like to retain it if possible. So, as an experiment, I've used prefers-contrast as a signal that I shouldn't do that. So now there's a bit of extra CSS under prefers-contrast: more which just uses the normal text colour for those headers.

I imagine this won't boost the score on PageSpeed Insights -- presumably it won't simulate that being used -- but I would hope it's an improvement for those who do need the extra contrast.

The other issue that was flagged up was the lack of an underline on the email address link in the comment-via-email invitation box (the optional feature I added back in v2.17.0). That one was an easy fix. There was no reason to avoid an underline on the link, so I added one.