It seems like it's the cheapest way to access Claude Sonnet 4.5, but the model distribution is clearly throttled compared to Claude Sonnet 4.5 on claude.ai.

That being said, I don't know why anyone would want to pay for LLM access anywhere else.

ChatGPT and claude.ai (free) and GitHub Copilot Pro ($100/yr) seem to be the best combination to me at the moment.

There's a lot of hate in this thread, but there are plenty of engineers chomping at the bit for autonomous workflows, because browser-use isn't there yet, and cloud expenses from major providers are also unappealing with so much relatively powerful local compute.

It’d be fine if they included a big disclaimer at the top that this is beta software and they’re not liable for blah blah blah, but without such a disclaimer it’s reasonable to assume the software is ready for production. I think much of the hate is coming from GH misrepresenting its software and people being surprised by the many minor bugs.

Thanks! We've moved the disclaimer up the page :)

You're absolutely joking if you don't think when this goes to crap Microsoft won't hoover up more share of OpenAI.

I wouldn't be surprised if they're banking long-term on its failure and got in early so they had equity preference.

It feels intellectually negligent to me that models like those used by Openpilot have no auditable concept of what a car, or a sign, or a person is.

Just where lines are and when a car should accelerate or break. The rest of the latent state is "based on pixels."

Your back-seat moderation is annoying and contributes to the wank of the site, too.

I wonder if legislation like this will bisect 3D printers before and after in the same way that ink printers use printer tracking dot steganography.

Or maybe 3D printers already do this in a way we don't see.

Apple has broken Time Machine enough times that I would never consider using it at all anymore. Once upon a time, it was really neat, had great integration with Mac OS X, and an amazing user interface and experience, but it's now clearly technology that Apple will probably eventually drop entirely in favor of something less impressive all together, like telling you to buy more iCloud Storage.

Hasn't the issues always been related to remote Time Machine? I have a usb drive I use and haven't heard of any issues with that setup. Am I missing something?

In the past, I've heard recommendations not to use remote Time Machine over SMB directly, but rather to create an APFS disk image on a remote server and then backup to that as if its an external hard drive.

Supposedly, doing that eliminates a lot of the flakiness specific to SMB Time Machine, and while I haven't tested it personally, I have used disk images over SMB on macOS Tahoe recently, and they actually work great (other than the normal underlying annoyances of SMB that everyone with a NAS is mostly used to at this point).

The new ASIF format for disk images added in Tahoe actually works very well for this sort of thing, and gives you the benefits of sparse bundle disk images without requiring specific support for them on the underlying file system.[1][2] As long as you're on a file system that supports sparse files (I think pretty much every currently used file system except FAT32, exFAT, and very old implementations of HFS+), you get almost native performance out of the disk image now. (Although, again, that's just fixing the disk image overhead, you still have to work around the usual SMB weirdness unless you can get another remote file system protocol working.)

[1]: https://eclecticlight.co/2025/06/12/macos-tahoe-brings-a-new...

[2]: https://eclecticlight.co/2025/09/17/should-you-use-tahoes-ne...

SMB on macOS is and always has, and probably will always be utter shit.

Mount something over NFS< and you'll be relieved about how snappy things remain. Snappy relatively of course.

Yes, there's some bug in the backupd that panic.. no matter smb/nfs

I tried moving to NFS, but the level of complexity of NFS auth is just comical. I gave up after trying to set up a Kerberos server on the Synology that I was trying to access. It's too much.

Using unauthenticated NFS, even on a local network, is too dodgy imo.

I lose my Time Machine drive, like, every year or two.

Sometimes, Time Machine just goes stupid and I have to wipe the drive and start over. All of my efforts in the past to copy or repair or do anything to a Time Machine drive has ended in folly, so when it starts acting up, I just wipe it and start anew.

Other times, it's the drive itself, and I swap it out.

99% of the time, it Just Works. Wiping the drive for me is more annoying than catastrophic (99.9999% of the time I don't care about my 18 month old data). It's mostly for local catastrophic fat fingering on my part, and to make sure I have a solid back up after I do a OS update. I have BackBlaze for "Why is there 5 feet mud in my burning house" scenarios.

Outside of that, I've always been able to recover from it.

My wife has a SSD drive she plugs into her laptop for TM backup. That machine at most makes laps around the house, so its not that big of a deal for her.

Replication isn't a backup. You need to make periodic copies of the sparse bundle / directory to actually have a backup.

Apple customers pay for backup solutions to backup data they don't care about and they don't even care when it fails.

The bar is so low!

Yes, the most important thing for apple’s customers is that they are able to pay apple.

I use remote time machine as seem to be fine.

There is an inverse relationship with respect to the size of an organization or project and its authors or maintainers willingness to review and merge pull requests.

The smaller a project, the more willing an author is to write the entire feature for you based on a plain request.

Go help smaller projects, the big ones don't care that you've submitted any work at all.

Linux is definitely not a "full operating system."

Here's Linux built on GitHub Actions, with Grub[1], and you can't do anything with it. I include a reference init that does nothing, per kernel.org. 17.8 MB image.

GNU is by every practical measure, everything else. People memed on Stallman for the whole GNU/Linux naming, but he's basically right. There's also Android/Linux, that another user mentioned, and some distributions which don't use a GNU userland at all.

But the wide majority of people are using GNU/Linux, or some ecosystem derivative of it, like people using GNOME, which was formerly a part of the GNU project.

[1]: https://github.com/andrewmcwatters/linux-workflow

Nobody cares! ”Linux” is used as a name of the OS.