Lemma 1: you want to protect your users privacy, and are also beholden to regulation enforcing that commitment (GDPR).
Lemma 2: you are obliged by other regulation to offer equal access to user data to third parties, so others can build equivalent functionality (DMA).
Lemma 3: malicious third parties will absolutely try to abuse the access and trick the user into sharing their data by all means possible. You will be held responsible in court of public opinion at minimum and legally at maximum if/when a malicious third party abuses said access.
This is a hard, possibly technically unsolvable problem no matter how much money you might have, because the root issue is not technical, it's the fact that you legally have to give third parties access and no way to control what they do with it - and as others have mentioned in the threads, it's exacerbated by the fact that the regulation doesn't say "this is okay and this is not", it is vague and judges things "by outcome", so you may spend all the time in the world implementing a solution you think will work, and then get hit by fines/lawsuits because the implementation is judged as not sufficient after the fact.
1: If you want to protect user's privacy, you collect no personal information, so GDPR doesn't apply.
2: You do that.
3: Since your platform collects no private information, they get nothing from you. If they collect private information on their own, it's their job to comply with GDPR.
What you should do in case (3) is ask the user for permission to allow the 3rd party access to private data on their device. It's their choice (not yours) to allow it or not.
I am not sure this is as much of a tension as you make it sound: where is the obligation that a marketplace administrator will be blamed for any and all breaches of data privacy trust from a participating (likely malicious) third party?
According to GDPR, the app developer is the "data controller" and thus ultimately responsible. Only in the case where Apple knowingly participated in unlawful behavior is it likely to be held accountable, and even then, in addition to the app developer. Obviously, if we are not talking about leaks from the actual App Store system (eg. Apple account logins and user data).
So while it sounds plausible, the legal framework is exactly not what you describe here — Apple can claim to want better protection for customers by not allowing third party apps, but EU rejects that (it can similarly extend to app store itself) and pushes for competitive landscape with DMA instead.
Like they now hand over all your contacts, your location, calendar entries, microphone access, camera access. If you choose to do so.
Nothing holds them from having designed this as an API that others can use where the user has permission toggles of what data they want to share with the LLM provider.
Interestingly, they claim to have done this and offer it as an API layer (Trusted System Agent) for other agents to use.
There is just this minor point that their own agent simply doesn't use it and goes directly to lower level interfaces nobody else gets access to: exactly the thing DMA was designed to stop.
This is clearly very different from usual permissions and access.
This would be unprecedented access to user data, enabling the most complete user profiling ever.
Ad companies, like Meta and Google, are going to spend huge amounts of money getting agents ready, because there will be a ridiculous amount of money behind all the data they're going slurp up, and the profiles they'll build for you.
Unless, Apple can figure out how to keep the leaches, that have consistently proven to be so, with court cases for receipts, at bay.
Very interesting - I just installed pg_hint_plan [0] extension a few months ago to get around a query that was confusing the planner too much. Edge case, but when you need it you really need it.
Time to revive my account... (scrobbling since 2003, heh) - started in audioscrobbler days, before last.fm
My usage has gone way down since switching to streaming services, this made me search for ways to backfill the data...
For anyone using Apple Music, looks like you can request an export of your data from privacy.apple.com, then use https://github.com/nerveband/Apple-Music-Play-History-Conver... to convert it for Last.fm import (haven't tried it yet, but looks promising).
Would be nice but it can't be done. I stopped scrobbling in 2011, then found out the site was still alive in 2024, one email to support later and I was back in.
I wanted to import my Spotify listening history from the intervening years, but found that the LFM API limits scrobbles to no earlier than two weeks ago. A shame, because my current listening history is now very skewed between the two "eras".
Yes, they are duplicates, although the links themselves are distinct. The current post points to my website, whereas the other points to my Git project. But they are indeed duplicates, in the sense that both refer to the same project.
Further, this post is from a new account with no posting history that has copied my submission (along with the exact wording) from <https://lobste.rs/s/hjipba/>. Copying the wording is no problem at all. I appreciate the visibility this post is bringing to this rather little project of mine. But I am not sure whether this post was made by a genuine user or by a bot. Any clarification from the poster or the moderators would be helpful.
I am also worried that one or both of these submissions may get flagged as duplicates. I had submitted the project myself earlier today at <https://news.ycombinator.com/item?id=47422759> but it did not gain any traction. I have emailed the moderators to ask whether it would be possible to merge the multiple stories into one.
I'll tell you what I specifically and intentionally do when I need to change lanes. I brake slightly, signal, and wait for the person on my right or my left to pull ahead of me, then change lanes immediately _behind_ them. Then sit there for a moment until my following distance evens out a bit.
This ensures that
a) I do not cut anyone off accidentally, and minimize the amount of stress in my immediate part of the universe
b) I will (most likely) have plenty of room behind me after I change lanes, reducing chances of anyone else running up on me
c) If there's noticeable traffic, the time I spend signaling and waiting for the person to move slightly ahead of me gives plenty of warning to the people _behind_ them that I'm about to enter the lane.
Ultimately, yes, of course in principle you're right, when I change lanes, I enter the lane in front of someone.... but I _can_ control whether I enter as far as possible ahead of them.
You shouldn't be braking when changing lanes is what I was taught, you should be matching the speed of the lane you're merging to. There are many drivers who think that braking is always the right solution, when sometimes it's a little more gas.
And in inclement conditions, it can make the difference between losing control of your vehicle or not. When you brake, you decrease your steering ability in most cars. Fine when its calm and sunny in CA, not so much when it's icing over near Ashland OR on the pass.
Well, sure - braking is mostly relevant when merging to the slower lane, when merging to faster lane I generally do not need to - since that lane is already moving faster, just need to speed up slightly and time it for the right moment.
My point is, it feels safer and easier to aim to enter a new lane with the aim of "following" someone, rather than trying to rush in "ahead" of someone. But maybe it's just me.
Since people are posting links to alternatives, another awesome source is the noun project. Has a mix of royalty-free, Creative Commons CC-BY-3.0, and paid license icons.
This is a very cool app and list of resources for learning Japanese. Does anyone know of similar top recommendations for learning Chinese (Mandarin or Cantonese?)
I believe Yomitan has Chinese dictionaries available for it but I don't know much about it. I would like to add Mandarin/Cantonese to Manabi Reader before long.
While I agree that the training/learning ecosystem is pretty heavily centered in Python, going from that to "Ruby is awful" seems like a very drastic jump, especially if we are talking about the LLM interaction only.
I probably wouldn't write a training system in Ruby (not because it's not doable, just because it's not a good use of time to rewrite stuff that is already available in python ecosystem)... but hooking up a Ruby system up to LLM's for interaction is eminently doable with very little effort.
I am assuming your situation had some specific constraints that made it harder, but it would be nice to understand what they were - right now your comment describes a more complicated solution and I am curious why you needed it.
Lemma 2: you are obliged by other regulation to offer equal access to user data to third parties, so others can build equivalent functionality (DMA).
Lemma 3: malicious third parties will absolutely try to abuse the access and trick the user into sharing their data by all means possible. You will be held responsible in court of public opinion at minimum and legally at maximum if/when a malicious third party abuses said access.
This is a hard, possibly technically unsolvable problem no matter how much money you might have, because the root issue is not technical, it's the fact that you legally have to give third parties access and no way to control what they do with it - and as others have mentioned in the threads, it's exacerbated by the fact that the regulation doesn't say "this is okay and this is not", it is vague and judges things "by outcome", so you may spend all the time in the world implementing a solution you think will work, and then get hit by fines/lawsuits because the implementation is judged as not sufficient after the fact.
reply