For context, there have been issues with MacOS 26 which have led many people to defer upgrading until MacOS 27 is available, and MacOS 15 is the previous version.
Also a liquid glass hater, but for what it's worth, 27 is supposed to make it a little bit better. I would prefer to go back to what I had before I had to upgrade into this horrible UI, but this is better than nothing.
I've been running the 27 beta and it fixes so much of the ugliness I hated in Tahoe. Lots of places had toolbars restored, solid areas that indicate where the window can be grabbed. The shitty icon spam in menus is back to sanity. And window corner radii are consistent and less bulbous.
macOS 17 is decisively faster on my MacBook Air M1 16 GB than macOS 26/27 is on my MacBook Pro M2 Max 32 GB. And I don’t mean just the animations themselves but lag/sluggishness/responsiveness in general.
You can already turn off most glass effects in 26 via "System Settings => Accessibility => Display => Reduce Transparency".
I had that turned off years ago (for reasons I don't remember), and was wondering what all the fuzz was about when 26 came out because I didn't see much of a difference ;)
IMHO the actual important visual changes in the 27 beta is that rolls back the bizarre oversized corner radius in Finder windows, and they also got rid of the 'every menu item must have an icon' idea.
Also on the beta. Agreed. Definite visual and UX improvement over 26, fixing the most egregious issues with 26 implementation of Liquid Glass, but maybe not quite as good as 25 overall.
As long as the UI improves and I can ignore all the AI stuff they're starting to push through, that's fine with me, though like many longtime macOS users, I'm not holding my breath for a bug-free experience.
A couple of weeks ago, I had a Baltimore Oriole (a cool-looking bird, not a baseball player) in my yard. They aren't rare, per se, but they are uncommon.
Took my iPhone out to snap a picture, and pressed the camera button. I hadn't used it, since upgrading to 26.
It takes the picture. It's there. I can see it, but it won't let me save it. Instead, it wants to tell me about the cool new voice-activated AI retouch feature. There was no way to save.
I probably could have figured it out, but I was so furious, I just nuked the picture.
for real? it did, like every subsequent photo taken on the device after pressing the shutter “button”. You must be a bot, just another living shill. another confident user error in the field
For those of us older than just 10 years of using macOS, the older Apple OSes have instilled within us the desire to never install the X.0 release and wait until at least the X.1 release. The bug-free experience is a myth
I found if you turn off transparency effects and turn on high contrast in the accessibility settings, it's actually a pretty nice looking UI. Still has some dumb quirks like inconsistent corner radiuses and text appearing under UI elements (because they're assumed to be transparent, but aren't). But all software is trash in 2026 so be thankful it isn't even worse, I guess.
Been doing that for years now already (performance boost). You can also do high contrast mode per app (if were talking about iOS). Must have for iMessage I think personally as it makes hard to read Green Bubbles a nice dark green that is very pleasant.
I bought a MBP mainly for the hardware, otherwise I'd have stuck with Linux. OS-wise, the jump from Monterey (which I'd last used) to Sequoia was smooth and still feels that way. I prefer not to notice the OS at all, something that I feel would be hard to do on Tahoe due to all that liquid glass and dumb transparency and animations.
Love to see this, now if only they could revert the whole shimmering oil slick disaster that is "glass" on iOS (unlike macOS, even if I "reduce transparency", the oil slick effect still appears in many places, in addition to all the spacing issues).
I recently made the move away from an out-of-support macOS 12 Monterey to Debian Stable (13 trixie), after some failed attempts to upgrade macOS using OpenCore Legacy Patcher.
While Linux support will always remain far from perfect with that series of MBP, it was still much better than I expected compared to my last look many years ago. I happen to get better use out of my particular setup now compared to macOS; as it's basically a glorified streaming device that I interact with through wayvnc/vncviewer out of arm's reach across a table. This also allows me to mostly avoid using the semi-busted Butterfly keyboard.
In addition to having something with security updates again, I'll now never need to think about Apple dropping Intel support. I'd also like to note that nix-darwin support on Intel macOS is ending very soon too - but not Linux + nix as a package manager (nixpkgs) since that combination is separate to macOS + nix-darwin.
Fortunately, those can run Linux. I recently installed Arch on an Intel T2. The only issue is that it does not have a TPM module, so the LUKS password needs to be manually entered at boot.
You can use TPM with secure boot to store the password. TPM checks that the firmware is the same and that the OS is signed by trusted (by _you_) keys, and if everything matches it makes the key available for reading by the OS.
Same. It's not the worst thing in the world, at least with Reduce Transparency enabled in the Accessibility settings, but I still don't feel any inclination to upgrade. My personal Mac Studio is macOS 15 and my work MacBook is on macOS 26, and I don't think there's a single thing that I find to be better on the work laptop than on my personal machine.
I might update to macOS 26 in September to be ready to update to macOS 27. Being two versions behind doesn't seem reasonable and I'd rather be on the "Tahoe but less shitty" version than Tahoe itself.
Same thing happens almost every release. I've stopped updating my Mac machine until I see something in the release notes I literally have to have in order to continue doing macOS/iOS builds, otherwise I'm staying on the version I've validated to work, and I know the existing bugs with.
That doesn't work any longer given patch tuesdays, at work wait that IT validates them and pushes the updates via managed WSU, at home, it is worthwhile wanting if something hits the news on WindowsCentral, Verge or what have you.
"You can delay this update for a little while" simply isn't the same thing as "you still have full control over updates".
It's right up there with Microsoft's "you are not allowed to turn telemetry all the way off", or all the efforts to require the use of an online Microsoft account to access your own computer.
Yeah, I thought so too, but surprise surprise; some months ago one of the "minor" updates "broke" ("upgraded") something that made my CI/CD setup stop working, that's when I dropped the idea that Apple even do "minor" updates anymore.
26.0 had a very annoying video jitter issue, but that was the first things that I noticed to be fixed in the next 26 release. Other than that, it worked just fine.
Did they un-hardcode the corner radius? I mean were they able to? I mean not that that anyone at this point needs convincing how utterly disgusting incompetent they’re at software.
I was shocked to find out the inconsistent corner radii in Tahoe was an intentional design decision. I figured it's so obviously bad that it was just sloppy work. But one of the WWDC session videos bafflingly clarified that it depends on whether the app has a toolbar or not.
I am fairly convinced this was all designed only in graphic tools, and looked nice in static mockups, like those fake desktops with overlapping small windows widely spaced, etc. The design folks probably all had giant studio displays with plenty of room, etc.
Then the reality of "what about toolbars", "what about dark mode", "what about laptop screens", etc were all afterthoughts and resulted in bolt-on fixes like that.
Map the amount of fixes with "... improved bounds checking...", "...improved memory handling...", "...improved memory management..." into the amount of developer, QA and release management teams salaries per hour, versus other stuff they could be working on, and that gives an approximate value of how using specific languages maps into monetary loss, and why companies are starting to care nowadays, given computers are always exposed to the world network.
>>, and that gives an approximate value of how using specific languages maps into monetary loss, and why companies are starting to care nowadays, given computers are always exposed to the world network.
You need also factor development time and ease of finding developers willing to work in a specific language. There are other factors like readability of the code (very verbose languages are likely to be worse) and cost of maintenance - languages forcing a lot of abstractions are likely much worse.
If you just meant apple-targeting developers, then yeah; you're right that those languages are in decline. But if you meant developers in general, I think you'd be surprised how many growth sectors are hiring C programmers. They're often not SaaS tech companies, but they are massive and many are growing. Hardware, industrial control systems, defense/aerospace ... there's a ton there, the spaces in which they hire just don't overlap a ton with the spaces frequented by hacker news. Also, a lot of them aren't US based companies.
I hope that changes over time, since I definitely agree that the downsides of C-family languages massively outweigh the downsides of competitor languages.
Which is why WG14 and WG21 caring about security would be quite relevant, but alas, priorities.
C could have gotten slices already in the 90's, the concept already existed in other languages, and even Dennis Ritchie made a fat pointer proposal into that sense.
The others, let see if anything related to profiles actually gets into C++29.
Citation needed. I don't think there's a correlation there. Over-architected Java spaghetti is verbose and unmaintainable. Under-architected Perl code golf that metastisized is terse and unmaintainable.
> languages forcing a lot of abstractions are likely much worse
Citation needed. C++ has had some very high-level abstractions on top of a low-level runtime for awhile, and plenty of people have decided to use it and hire for it regardless. What counts as an "abstraction" or "forced abstraction" is a very very subjective topic.
Even C can be super abstracted, see early 1980-90's business software written in C, with nice stuff like Yourdon Structured Method, leading to over-architected C spaghetti with macros, a decade before Java came to be.
The problem is the lack of interest since Morris worm came to be, to provide better mechanisms in said languages, until governments and key big tech names decided it was time to change existing practices.
It was my vague understanding that by the time you implemented all the apis needed to run normal software on top of that, you either have enough apis that different tasks can still compromise each other, or you have shoved everything into a single task with very little isolation between normal user processes. In either case, it doesn't seem like you actually gained so much. What am I missing?
Lots of "in collaboration with Claude and Anthropic Research" mentions, no mentions of other labs. I'd assume Apple already had access to whatever the most powerful model is at the various US-based labs, but perhaps not?
Those were voluntary disclosures by two Anthropic researchers and the security firm Calif. I know one more CVE on the list that was discovered using an AI agent and wasn't disclosed as such. I suspect there are many more.
The gist is, OpenAI hired a high ranking Apple employee who helped other Apple employees get hired by OpenAI and exfiltrate Apple trade secrets in the process.
Given that his departure was marked by Apple products getting more reliable and usable, they arguably did too much in that regard (one of Cook’s more notable bad calls). Once he stopped blocking it, the keyboards were fixed and pro devices regained enough ports for pro users.
Without his support, his protege Alan Dye left for Meta and improved the design skills at both companies.
> Lots of "in collaboration with Claude and Anthropic Research" mentions
I wouldn't say 4 is lots. The entire list is massive. I haven't counted myself, but someone claimed that macOS 26.6 has the all-time record with 155 CVEs.
Considering that in Feb 2026 (https://support.apple.com/en-us/126348) Claude wasn't mentioned even once, 4 sure sounds like "lots" compared to nothing :) But you're right, it's subjective ultimately.
Do they? As in Claude but on premises? Wonder if this is gonna be the solution that e.g. banks will require, exactly like they do now for cloud services (e.g. Azure on premises).
Pretty extreme solution… you can get Claude models from AWS Bedrock and Google Model Zoo. These are both very helpful for compliance and security, but do require you to have a cloud strategy.
I don't believe it's published anywhere, but it's common knowledge to Apple engineers. I can second the poster's assertion that they run Claude internally.
I don’t know about hosting it internally but a an Apple focused podcast I listen to (Accidental Tech Podcast) has mentioned numerous times in the last few months they’re using Claude heavily. And they know Apple insiders.
It is a lot of fixes and the Android Security Bulletins of June and Android 17 also had a lot of fixes [1], despite ASBs only containing high/critical vulnerabilities (other vulnerabilities are only fixed in major releases and QPRs, which most Android vendors respectively roll out late or never at all).
I think the story here is that vulnerability discovery has accelerated a lot with LLMs, but since are adversaries are doing the same, it is more important than ever to update quickly (and not let some Android vendors get away with their lazy update schedules).
Right - it was always dangerous but people who figured they weren’t important enough to be attacked might find out that LLMs have shifted that cost in the wrong direction.
And it's not actually that much information "about the security content". For example: "Impact: An app may be able to access sensitive user data. Description: An access issue was addressed with additional sandbox restrictions." This references CVE-2026-43819, which doesn't have any more information. Compare this with the nearly decade-old https://support.apple.com/en-gb/103680, and you see much more specific information about problems and their remedies (except in situations where Apple's action was to update a vendor component).
The vagueness could be intentional. There’s been a big issue with linux where proof of concept exploit code gets posted before the bug is announced because people reverse engineer it from the fix commits.
Apple has the advantage that they can keep everything secret for long enough for the patches to roll out. And realistically there is no reason the user needs to know the details of an exploit that was patched before it was ever used.
Remember that they have a great deal of telemetry around things like crashes and work with groups like Citizen Lab for certain high-risk users. You can’t prove that something was never used in a perfectly targeted and concealed attack but it’s likely they can say it wasn’t used outside of such contexts, and once you’re at the level of things like “the Mossad deployed an exploit after configuring the local cell tower to drop external network access before crash reporter could phone home” user notifications in the release notes aren’t effective anyway.
This may be a naive take, so if anyone has insight please feel free to share, but across Windows, Mac, and Linux OS's I see many cases of path parsing vulnerabilities resulting in sandbox escapes, code execution, or data access issues. When presenting the user with a file picker or command-line input, is it really needed that the software can handle the full POSIX spec?
I do not see a "typical" user needing to access a path with say a network storage but multiple ../.. and hard and soft symlinks simultaneously.
I think "be liberal in what you accept" might need to be revisited for path parsing with some sort of OS-wide single-implementation as an optional feature.
What I mean is that for “honest” software, built-in to the OS or otherwise, the programmer finds a situation where they take some user-supplied input and concatenate that into a path, and call something like OS.read(). If they want to prevent the user from causing havoc, they now find themselves dealing with path validation in their software instead of calling OS.safeOpen(), which would be a reduced subset of allowed chars?
How is it permission denied if the app is running as admin? The OS "handled" it by giving admin app admin access. Sure, it was tricked by the user input, but that's what the fixes are for?
Collision counts are absurd. CVE-2026-43739 has roughly twenty credited researchers; CVE-2026-43816 has nearly as many. And ai attribution getting credit.
Apropos, anyone else saw "fast user switching" in Tahoe turn into "excruciatingly slow user switching which after a minute of switching without success rebooted the whole damn machine"?
For context, there have been issues with MacOS 26 which have led many people to defer upgrading until MacOS 27 is available, and MacOS 15 is the previous version.
reply