There's a company, currently called Tie (meettie.com), formerly known as Revenue Roll, who promises to "de-anonymize your highest value web traffic", which in practice means that they give you an email address for retargeting, for a user who visited your site without ever explicitly providing any identifying info.

The old site had a blog post [0] where they explicitly said they were using fingerprinting, and even called it "privacy-compliant".

I'm sure they're not unique in the service they provide, but that was the first time I'd seen someone brag about browser fingerprinting.

[0] https://web.archive.org/web/20240527125312/https://www.reven...

I notice they have an Opt-Out form here: https://app.termly.io/dsar/ee5088c4-5eb2-475c-a9ea-9376f1b70...

It's pretty hilarious legalese and tells you nothing about what it even achieves. Maybe makes you a Very Important Marketing Target.

One thing that struck me was the 'Under penalty of perjury, I declare all the above information to be true and accurate'. Shame they seem to require validating request by email. It'd be fun to take a PII breach and throw all the emails you find at 'em.

Why aren't privacy orgs trying these services.

Messenger is also the only one of Meta's apps that I still use regularly and I was pleasantly surprised with their decision to roll out E2E encryption by default.

https://about.fb.com/news/2023/12/default-end-to-end-encrypt...

I've never signed up for the X developer program, so I'm not bound by these terms. But I did download an archive of my data last week. Do I have implicit permission to use that data (~150k liked tweets) to train AI models?

Or is there stuff in the user agreement that separately prohibits this?

Obviously barring normal copyright law which is still up in the air.

If you live in the EU, GDPR dictates that you own your data generally speaking. If you're in the US it varies by state if you have any rights at all.

If you own your face that doesn't mean nobody can take a picture on the street.

That is partly true, there are carve outs for public spaces. But what they can do with that image is still governed by the GDPR in the EU.

>> but you have no idea what the code is

> Are you a vibe coding Youtuber? Can you not read code? If so: astute point. Otherwise: what the fuck is wrong with you?

> You’ve always been responsible for what you merge to main. You were five years go. And you are tomorrow, whether or not you use an LLM.

Loved this bit. In a professional setting, any code — AI-written or not — should be read by at least two people, the submitter and the reviewer.

What exactly is the advantage of Swift UI over Flutter? Maybe it's slightly more efficient since Flutter does its own rendering, but in my experience I've never run into issues with performance.

And I think everything should be web apps anyway (ideally PWAs), but I like that Flutter lets you produce a desktop app from your mobile app with very little effort. Even without any special "integration" with the OS, it's better than packaging a web app in Electron, right?

Our app is written in SwiftUI because of the ease between iOS, iPadOS, macOS and Apple Vision. There is just minimal configs to make it work between them. I don't this can be done with any cross-platforms.

Flutter doesn't really need any special configs either. Going by platforms, I think most developers would rather target Windows + Android + Linux over Apple Vision.

I don't think that's true. There are a lot of people on Twitter who keep accidentally clicking that annoying button that Elon attached to every single tweet.

Cool game!

Ran into a weird bug on puzzle 2 where adding a `flex-flow` property and then deleting it also seems to unset the `flex-direction` property on the underlying element. Probably should be disallowed under the no-mixing-shorthand rule I guess.

Also could you make it so that clicking inside a text box doesn't trigger the drag effect? It would be nice to be able to highlight/double-click text to delete/replace it.

I didn't read JS The Good Parts until it was well outdated, and I was glad to see that a lot of the sharp edges that Crockford lists have largely been eliminated. The book was written circa ES3, so some problematic features were removed in strict mode, we have replacements for some (let/const, for-of loops, etc), and we can sweep prototypes under the rug with ES6 classes, sometimes arrow functions even avoid awkward this-binding issues. The rest, like you said, TypeScript + a linter takes care of.

The images are not destroyed, they all still exist.

I'm not sure if a style is something that can be destroyed, but I don't think AI has done that yet.

I used to be a fan of Star Wars. Then Disney bought the franchise, and transformed it into something that a large number of fans, including me, did not like.

That the old movies exist* does not redeem the current franchise for me.

*: albeit in reality the original films are being preserved by fans and not by the official rights holder.

It's all relative. I thought Lucas let Star Wars wither for too long, extended universe not withstanding. Most of the Disney works based on it have pleased my family and I.

Styles can absolutely become wack, if the context they're used in becomes wack.

You're right that quantization isn't anything special here, but red herring isn't the right word, it's just "embarrassingly simple", per the title.

Okay, but narrowly focusing on a "quantization-robust unlearning strategy" as per the abstract might be a red herring, if that strategy doesn't incidentally also address other ways to undo the unlearning.

I think it's useful because many people consume quantized models (most models that fit in your laptop will be quantized and not because people want to uncensor or un-unlearn anything). If you're training a model it makes sense to make the unlearning at least robust to this very common procedure.

This reminds of this very interesting paper [1] that finds that it's fairly "easy" to uncensor a model (modify it's refusal thingy)

[1] https://www.reddit.com/r/LocalLLaMA/comments/1cerqd8/refusal...

Yeah, exactly this. You would really want to pursue orthogonal methods for robust unlearning, so that you can still use quantization to check that the other methods worked.