Lazare

Yes. Cursor marketplace is a mirror of Open VSX except they inject some plugins that are not in Open VSX for some reason. This works because Open VSX has basically no security. Which was already an issue.

Anthropic wasn't hacked. Cursor was.

Cursor maintains a mirror of OpenVSX... Kinda. Except they inject some of their own extensions in it. But it's mostly a mirror.

Edit: Cursor is apparently stating now that the tampered files were only on their mirror not OpenVSX.

Cursor has their own mirror of OpenVSX. Some reporting suggests the tampered files were on the Cursor mirror but not the main OpenVSX. (This is shown, among other things, by the Anysphere plugins available in the Cursor mirror but not in OpenVSX.)

Edit: It seems Cursor is claiming the tampered files were only in their mirror.

As far as I'm following it, the court didn't suggest the prompt was not a protected work.

But what the court did say is that the logo wasn't (sufficiently) derived from the prompt; those 300 words weren't contained or encoded in the logo, so if you copied the logo without permission from the author of the prompt, this was fine, because you weren't in any meaningful sense copying that prompt, or any other protected work of the author. (Which in my view is fair enough, really!)

Another way of thinking about it is to consider:

1) The number of different prompts that would yield a very similar looking logo. In many cases, a one or two word difference would probably yield a very different logo. (Conversely, you could rewrite an entire chapter of the novel, but the other chapters would still be translated more-or-less unchanged. You'd need to rewrite every sentence to end up with an entirely new set of sentences in the output, and since copyright covers the arrangement of text, not the plot or ideas, that matters.)

2) How feasible it'd be to look at the logo and accurately guess the prompt. (Again considering the hypothetical novel, if you translated the novel back into English, entire passages would likely be word-for-word identical with the original.)

Interesting thought experiment is to consider an author who writes a novel in English, but wants to translate it into German.

They have a copyright on the original, and if they hire a human, the human would have a copyright on the translation (which would generally be licensed or transferred back to the author in some way).

If they use an AI for the translation, by the logic here, the translation wouldn't have its own independent copyright, but (based on other long established principles of copyright) it would still be a derived work of the original, so even if this decision holds it would not be legal to make unauthorised AI translations, pirate authorised AI translations, make further translations into other languages (or back to English), etc.

Which seems fairly reasonable! But consider:

If you start with, say, a 90,000 word novel, and ask for a translated novel, you (presumably) have sufficient rights to stop someone making unauthorised copies of the AI translated version.

If you start with a 300 word prompt, and ask for a logo, you (apparently) do not have sufficient rights to stop someone from using it without authorisation.

So some combination of the input (0.3k vs 90k) and the output (logo versus novel) crosses an inflection point between these two extremes, and I think it's interesting to wonder what the boundaries are. Like, in theory you could graph input size versus output complexity, and sketch a frontier between "the author's protected expression survives in the output" and "the author's protected expression does not survive in the output". And I don't have the slightest idea what I think a fair frontier would look like.

That is the claim, yes.

But no evidence has been presented, and it feels very unlikely to me.

I'm a bit skeptical Google ever treated `[email protected]` and `[email protected]` as different accounts.

I registered the `fname.lname` version very early on (before GMail was freely available), and while I do sometimes get emails sent to the `fnamelname` variant, which is consistent with someone having signed up with that variant, but is also consistent with someone just forgetting what their email address is (eg, they meant to type `[email protected]`, or `fnamejlname`, or whatever). Especially since the mixups are very, veyr rare.

I can't rule out some weird collection of google bugs, but Occam's Razor suggests it's just an occasional typo. Especially until someone pulls up two screenshots of Google accounts showing colliding addresses.

Unfortunately, that's unrelated; only models released after August 2nd support it so far, according to Anthropic.

As per the link, the words in the green and red groups are calculated dynamically, so it's not like the model is going to be told "use 'unique' over 'unusual'" and suddenly writing from the model will contain the word 'unique' far more often than 'unusual'. So I'm not sure it's clear that this has an impact on creativity as such?

That being said, I do question how this will apply to code as opposed to prose. Even data dense text (ie, if you ask Claude to evaluate what running shoe to buy, and it spits back a list of options with reviews and prices) may struggle.

What it probably will work well at it flagging the current tsunami of entirely AI generated novels on Amazon/Kindle, which is...honestly not without value.

> Identifying code will be paramount in training future models

True, but note that this strictly allows providers to identify text generated by their own models. If Anthropic wants to filter out GPT generated text in their training data, they'll need to feed it through an OpenAI API, which is implausible. So it might help on the margins, but I don't think it solves the problem of model collapse.

Headline discusses missiles (which are topical); the article is mostly about radars in F-35s, and admits halfway through it's because a procurement schedule slipped.

This is just clickbait; there's no content here.

Bets with a positive edge can be found most commonly via sports betting, poker games, and the stock market. Although note that most bets you can find via such methods do not have a positive edge, and the edge is not labelled.

More generally, look for any area where bets can be made, and professionals exist who make their living winning such bets. There are professional poker players and sports gamblers; there are not professional roulette players.

Not exactly on topic, but I have to endorse mise; it's a really, really great tool that solves a number of different problems.

Example: I was recently working on a large project that needed a specific version of Python, and there's a lot of ways to solve that, but mise was an easy and robust one. But also, the project needed a bunch of different tools to build it, deploy it, do local dev, perform certain maintenance tasks like rotating secrets, work through certain operational runbooks, etc., and mise was an easy and robust way to solve that too. Once you know everyone on the team will have the same tools available, if a runbook would be simpler if you could assume everyone has jq installed, well, just add it your the project's mise config, and now they do. And then when I switched to working on a Java service, and then later a Node service, well, obviously mise was an easy and robust solution there too.

By contrast, I made an effort a year or two back to adopt nix, which (despite starting from a very different place) solves a lot of similar problems, but found it a bit daunting (large, complex, poorly documented, and felt hard to partially adopt), and while I love the concept of nix, as a practical matter I ended up abandoning the effort. But mise was really easy to understand, adopt, and progressively add to an existing project without unduly impacting other team members. (Example: Mise will read existing verion manager configs from tools like sdkman, which makes adopting it over time easier.)

It's got to the point now where I'm using mise in place of Homebrew or other system level package managers for basically all CLI tools. Which feels weird when I think about it, but mise genuinely just feels like the better solution. If mise has a flaw, I haven't stubbed my toe on it yet.

> Jani Patokallio runs gyrovague.net in order to harass people who provide useful public services.

I mean...investigating who runs secretive yet popular websites is a useful public service, generally called "journalism". And your comments in this thread could be seen as an attempt to harass Jani.

I do not, to be clear, think you're doing anything morally wrong, but I'm also not sure I see how you can draw a bright line between your actions and Jani's. By the rather stretched logic and loose standards you've been using in these comments, it seems like you run your HN account to harass people who provide useful public services, no?

I'm absolutely open to the argument that Jani has does something wrong, but nothing you've said has really even accused them of anything.

If you want to define doxing narrowly (as it was historically) then I would agree that all (or nearly all) such cases are wrong, but this is by no means clearly doxing. If you want to define doxing widely (as is common lately), then I'll accept this is clearly an example of doxing, but note that there's nothing inherently wrong with doxing.

Just saying "doxing" does not establish that the underlying actions are immoral, and so it does not follow that the target not appreciating it is relevant. If I take the last parking place in a crowded lot, the driver behind me certainly won't appreciate it, but I have no obligation to give it up. If you think Jani has done something incomparable to taking a parking place, you need to make the case.