I think I'd prefer a pane view over a tree view because the higher levels retain their scroll position even as I browse beneath. Is there a similar UX possible for tree views?
Until something happens that disproves this, my personal belief is that supervision and manual review is one of the best ways to use a coding agent. You don't need to understand everything it does, but you will benefit from a technical background and from at least surface-level knowledge or intuition about what it spits out.
I review every diff Claude Code applies and periodically re-review entire projects as a whole. Through this, I've managed to keep architectures fairly principled with future expansions in mind. I recently managed to essentially two-shot an MLX implementation of a working forward pass for a diffusion language model, based on CUDA source code that is not compatible with my machine. There's more work needed before it's anywhere usable in practice, but the fact that the model is now running at all on my machine is a very impressive start.
For that, I had it study the CUDA source code and write a very detailed document with its analysis of exactly how the model is implemented. This document only had one material flaw. Then it studied MLX for a while and spat out a running forward pass based on the flawed document. The output wasn't of sufficient quality, so I had it insert debug prints throughout the whole inference process to see where it was going wrong. It found and fixed the forward pass and the flaw in the document. I needed no domain experience in LLMs or DLMs for this (although I benefit from some minor past contributions to RWKV.cpp).
Another example is that I recently started getting into SwiftUI, and Claude Code is doing a very good job at demonstrating code patterns and pointing me towards APIs that may solve my problems. It also helped me set up things like API clients (which itself, of course, gave me pointers to all the sorts of documentation I'd benefit from reading in full). I reject a very large fraction of its suggestions, I tweak its plans very frequently and I tell it off a lot from things that either are unidiomatic or are objectively terrible hacks. But it is incredibly useful for menial work, for enumerating possibilities, for quickly scaffolding placeholder content, and for demonstrating patterns I haven't learned yet as they apply to my specific situation. For example, Claude Code quickly learned me how to use NSViewRepresentable, whereas in a past project where I didn't use LLMs, I absolutely struggled to embed a Metal view.
But all that is to say that I'm skeptical of solutions that try to have you describe your idea in plain language; that try to insulate you from the code; or that make the lie that you just don't have to worry about it. If you work at all on the kinds of projects I work on, which are chock-full of reverse engineering and an obsessive focus on tightly-controlled design and idiomatic code, I firmly believe that treating the code like a black-box is not the way to do it. I don't know if Glaze truly hides the code, but I don't see any mention of it in the trailer video and that makes me feel a little dismissive.
To an extent, logs like this are incredibly personal - or at least I'd consider them such - so I'd understand if they're not being released publicly for many reasons relating to that.
The kind of vulnerability that shows when someone is susceptible to influences like this isn't exactly the kind of thing you'd want to widely publicize about someone you loved, you know?
Google will not release the chat logs publically, it's up to the court how to handle them, but the bar for "the public cannot see this" is generally much higher than "well that's embarrassing". If this goes to trial, they will most likely become available.
Remember that the idea of the court is to be public and transparent, with judgement coming from the jury, but also to be judged by society on the whole. So if you're gonna sue your kink provider, be prepared for everyone to know how you get off, because after all, the court is owned by, and serves, the public.
I like to think of this like tempering the output space. With a temperature of zero, there is only one possible output and it may be completely wrong. With even a low temperature, you drastically increase the chances that the output space contains a correct answer, through containing multiple responses rather than only one.
I wonder if determinism will be less harmful to diffusion models because they perform multiple iterations over the response rather than having only a single shot at each position that lacks lookahead. I'm looking forward to finding out and have been playing with a diffusion model locally for a few days.
Yup. I think of it as how off the rails do you want to explore?
For creative things or exploratory reasoning, a temperature of 0.8 lends us to all sorts of excursions down the rabbit hole. However, when coding and needing something precise, a temperature of 0.2 is what I use. If I don’t like the output, I’ll rephrase or add context.
As the user interface through which users download (among other things) apps... it absolutely is an "app store". It's not where the binaries are hosted, but you don't see anyone claiming the App Store iOS app isn't an app store because the apps are ackshyually on Apple's CDN servers, do you?
Ok. Take something like Gnome software or KDE discover, where you can add different sources. You can point to both your distro's repos AND Flathub. Is the store the app? Is it Flathub? Theres different sources depending on which distro you use and users can tweak it.
Apple App store is a disingenuous example because it's a proprietary app hard-coded to use Apple sources, you can tweak the sources... Apt or yum are no more app stores than curl or git.
The app is a store, and Flathub is also a store if it allows downloading and installing the packages directly.
Apple's App Store is a perfect example because there is no difference between stores with private sources and stores with open sources for the purposes of whether or not it is a store.
Monitoring children's DMs is the responsibility of the parents, not megacorps. If a parent wants to install a keylogger or screen recorder on their child's PC, that's their decision. But Google should not be able to. Neither should... literally anyone else except maybe an employer on a work-provided device.
> Monitoring children's DMs is the responsibility of the parents, not megacorps
Absolutely. But what responsibilities do megacorps have? Right now, everyone seems to avoid this question, and make do with megacorps not being responsible. This means: "we'll allow megacorps to be as they are and not take any responsibilities for the effects they cause to society". Instead of them taking responsibilities, we're collecting everyone's data and calling it a day by banning children from social networks... and this is because there are many interests involved (not related to child development and safety).
Human operators were not required of The Bell Telephone Company by law. Bell switched to mechanical switching stations as soon as doing so was economically advantageous.
(Reconsider my post. I'm arguing for no regulation.)
I'd say that at minimum social networks need to be required to show how their algorithm works and allow users control over their data. They must be able to know why a content was served to them. Nowadays social networks are so pervasive in society, affecting it and molding it to unknown interests, that this is the bare minimum for a free society.
Ideally, users should be able to modify the algorithm, so they can get just what they want, while simultaneously maximizing free speech. If something isn't illegal, it shouldn't be hidden or removed.
> Nowadays social networks are so pervasive in society, affecting it and molding it to unknown interests
I think this is the real issue. We should free ourselves from "social networks" such as Tiktok, Facebook, Instagram and others. Even with direct messages truly E2EE, they create countless other privacy problems. They enable surveillance of people at scale and should be completely shunned for that reason alone.
> nor the complete mapping of customers behaviors to use in contexts not being the phone call
This is because the telephone system was regulated with wiretapping laws, among others.
> I don’t remember reading about ads in phone calls
See above, but also: Junk Faxes & Telemarketing/Robocalls.
> The apples to oranges in this comparison is probably top five on HN ever.
It all comes down to whether you view social media as a communications platform or a publishing platform.
The strict regulations governing the telephone system (see above) would adequately cover both with deliciously-antique turns of phrase like "wiretap" and "pen register".
Posts made are like the letters page in my view, but even if you don't believe that should be controlled, when it comes to the adverts these publish surely it has to be the latter.
Personally I argue they are more like a newspaper as they are providing a platform. Your ISP is more like a phone company or postal service as they just transport packets from one person to another.
Cloudflare and similar are more arguable, but to my mind if you host a computer you are responsible for the contents stored on that computer.
They should have a responsibility of transparency, accountability and empathy towards users. They should work for the user and in the interests of the user. But multiple constraints make this impossible in practice.
> Monitoring children's DMs is the responsibility of the parents, not megacorps.
Yup, but the tools provided make that easy or hard.
But putting that emotive bit to one side, Megacorps have a vested interest in not being responsible to children. They need children's eye balls to drive advertising revenue. If that means sending them corrosive shit, then so be it.
Its a bigger issue than encryption, its editorial choice.
I'm all for helping parents to do this. Any site requiring age verification should indicate this as a http header or whatever, and the browser I allow my child to use should respect that and the parental controls should be easy for me to engage with
Many parental controls are massive pains to get working. Apple does fairly well (although I don't get a parental pin number to unlock the phone, which is normally fine as my child will tell me, but in some circumstances it wouldn't be), but does require the parent to be on the apple ecosystem too.
EA and Microsoft however are terrible, especially as it's likely the child will be playing fortnite/minecraft and the parent won't have ever touched it. I think with minecraft we had to make something like 5 or 6 accounts across three different sites to allow online minecraft play from a nintendo switch.
Why? Plenty of children benefit from talking to other people. Some children need careful monitoring, and some children shouldn't be allowed to use DMs, but it's not universal and should be up to the parents.
Control over who they can talk to (if needed), certainly monitoring of both who they talk to and in many situations what the contents are
At some point between the age of 0 and 18 the child has to be fully ready for an independent world. A cliff edge is a terrible idea, allowing 3 year olds unmonitored uncontrolled conversations with strangers is a terrible idea, but not allowing 15 year olds to talk to their friends is a terrible idea.
Do we know if there there be Widevine L1 keys that aren't deleted on unlock? (Certain phones restore access to L1 on bootloader relock, as long as AVB passes, including with custom keys.)
reply