HN user
So websites can now nag users to allow access to the root of their local disk
No, Chrome doesn't allow this.
Here's a simple demo: https://output.jsbin.com/kekekac/quiet - note that you can't select root, Downloads etc.
The user agent is the proper channel for the agency Jake is seeking here. Theres nothing preventing the user & their user agent from negotiating what model they use.
This isn't how it works. As the developer, you use the system prompt to set a particular personality for the chat bot. Eg, when you use an LLM in VSCode, it comes with a system prompt to make it an effective code assistant.
Now, in VSCode, you can select a different model, which is maybe where your misconception comes from. But when you select a different model, it will also use a different system prompt, designed to achieve the same personality, but tailored for that particular model.
Once you figure out why they do that, you'll understand why your position here doesn't make sense.
Yes! It might lie or hallucinate. But also, all browsers claim to be "Mozilla/5.0" in their user agent string. It's a very similar problem.
It's fun that I get to be called both "anti-AI" and an "AI shill" by people on the internet depending on the day of the week.
Thank you for posting this.
On interoperability, time will tell I guess. I've only been working on Firefox for a few months, but general interop issues are way worse than I realised when we worked together at Chrome. Firefox frequently gets bug reports for not behaving like Chrome, even when Firefox is complying with the spec, and Chrome is not. We end up having to just behave like Chrome.
On developer signals… I'm sure there's better evidence of positive sentiment than Chrome provided, but there's a lot of negative sentiment too. I think it would be fair to call the developer signal "mixed", or maybe even "polarised".
Aww thanks for saying that! I've been doing little videos on https://www.youtube.com/@FirefoxWebDevs (and accounts of the same name, pretty much everywhere). Although they're designed to be short, so they're pretty different to HTTP203.
Agreed. Maybe Google will propose a CSS text formatting property that cannot be used on paragraphs that are critical of the US administration.
Like, that sounds daft, but it's not really far from what they're doing here.
I think the objection here is unrelated to the love or hate of LLMs. It's about the viability of this particular proposed open web API.
I personally use LLMs for coding assistance, and some home automation stuff, but I do not think this particular API is good for the web.
Aww thanks! To be fair I didn't toe the party line when I was at Google (imo). Although, that caused me increasing amount of grief internally, until I left. From what I hear, things have gotten exponentially worse in that regard for folks still on the team.
Sigh, when I posted this, I linked to https://github.com/mozilla/standards-positions/issues/1213#i... (which was posted 11 hours ago). Unfortunately someone changed the link.
When I posted this, I linked to the latest statement https://github.com/mozilla/standards-positions/issues/1213#i..., which is the content relevant to the title (the details of our opposition to the API). Unfortunately someone removed the link to the specific post.
The Chrome model requires either "16 GB of RAM or more and 4 CPU cores or more" or "Strictly more than 4 GB of VRAM", and "22 GB of free space" (it uses around 4.4GB but it doesn't want to use the remaining free space).
The model is pretty slow on my M4 Pro mac.
The API allows the browser to use a cloud service instead, but then privacy is lower. So, more privacy for the rich.
The rate of model development is an issue here. Once there are many cross-origin models, it becomes a fingerprinting vector. Also even the small models are many GBs.
A lot. My scores:
- 0.0028 on my MacBook pro screen
- 0.0045 on my Dell monitor
- 0.0033 on my Pixel 10 pro
And those scores are pretty consistent.
The associated deep-dive article is great https://www.keithcirkel.co.uk/too-much-color/
fwiw, if you serve your page with:
Content-Security-Policy: require-trusted-types-for 'script'
…then it blocks you from passing regular strings to the methods that don't sanitize.
Fwiw, JPEG XL takes around 2.5x the time to decode as an equivalent AVIF, and has worse compression https://jakearchibald.com/2025/present-and-future-of-progres...
I know this isn't quite what you mean, but when you first hit the page, the list is in a random order, but it's then stable across reloads.
I considered the 'vs' approach, but I worried that there might be a lot of iterations where one or two of the options would be things that the person didn't understand, or didn't care about.
How do you feel about something like this: The user goes through the long list, picking what they understand and don't dislike, then the 'vs' system is there for helping determine the order of those items. Then the user gets the ranking which they can tweak.
Author of the app here. This was quickly thrown together, and something we'd like to improve on next year. Any & all feedback is welcome.
We've already been looking at the results, and it has influenced the direction of multiple vendors in the process, but of course it isn't the only thing we look at.
We're going to continue to look at this throughout the process, so it's still worth getting a list together.
Whether we can release the raw results of this (and to what degree), will involve discussion between browser vendors, so it isn't something we can commit to right now. I know that's not… great… but it's a delicate process (one that I'm still personally getting used to - it's my first year 'on the inside').
Author of the app here. This was quickly thrown together, and something we'd like to improve on next year. Any & all feedback is welcome.
Browsers don't want to add new ways of running script. That said, I wonder if `<?xml-stylesheet type="text/javascript" href="script.js"?>` could work. It's kinda weird, but `xml-stylesheet` can already run script via XSLT, so it isn't a new way to run script.
Alpha channel support in x265 is very interesting, as this was only previously possible with paid-for Apple software (and the resulting file sizes were high). Some details from when I last looked at it https://jakearchibald.com/2024/video-with-transparency/#enco...
That document order thing doesn't sound right to me. Here's a demo where the popover appears before the anchor https://codepen.io/jaffathecake/pen/MYargba?editors=1100
`position-anchor` is a high-level simple way of doing it, and it comes with the restrictions you mention. However, the `anchor()` function, which is also mentioned in the article, gives you the kind of flexibility you want.
I guess you're being downvoted as a general nay-sayer, but you're right. I tried this feature last month and a bunch of browser bugs and design issues got in the way. I reported them, and they're being worked on https://github.com/w3c/csswg-drafts/issues/12466
The `margin:0` issue was particularly frustrating & imo should have been covered in the article, as it's a real gotcha when trying to use popover & anchor positioning in combination.
odd that the original post didnt have it either
Totally honestly, it was a lot of work, I got bored, and it was diminishing returns for the later parts.
I linked to it without realising that. I didn't know Aaron personally, but I remember people speaking fondly of him.
Right, but it should be easy for the reader to decide whether to divert to the additional content. A superscript number tells you nothing about the additional content. Whereas on the web, link text or a details summary can help you make a decision.
There are absolutely accessibility issues https://adrianroselli.com/2022/09/brief-note-on-super-and-su...
It's ok, but you're still being taken out of the flow. Especially as the notes appear arbitrarily on the left or right. And yeah, you have to find your place again afterwards. It's way better than scrolling solutions though.
That particular example on mobile falls back to a revealing pattern, which is pretty good.
It still has the problem where it's just a test of your curiosity. You don't really know what the supplementary content covers until you expand it. The link text is just a superscript number which is kinda useless.
This is why I prefer the solutions in the article where the supplementary content has a heading that hints at the content.