but there is a difference between deterministic compilation and non-deterministic LLM code. Of course I don't think this is an issue for toy problems and simple codebases, but for non-trivial problems I think it will be an issue.
When I compile C code I know that maybe it will not be as efficient as it could be if I had written it in Assembly, but there will be a biunivocal correspondence between C and Assembly. If instead I use an LLM to rewrite a feature of a codebase I can't be sure that it still functions like the original one.
I acknowledge that this is an issue with human programmers too, but I don't see a clear way forward, even if I'm really interested in LLM compilers being a thing. Maybe we will use them for non important code, and we will keep writing system critical stuff by hand.
> but there is a difference between deterministic compilation and non-deterministic LLM code
But not the point of my comment. Computer programming has been a progression of physically wiring up valves, to soldering transistors, to punched cards, assembly, then higher level languages. Now we have natural language models.
The analogy being each that most people don't care about the assembly generated as the code works and it's really performant/efficient. Humans can still optimise assembly, but there's vanishingly small marginal gains for all but the most intensive/low-level tasks.
If LLMs produce things that work, and are indistinguishable from a careful human programmer (i.e. with some level of acceptable performance), people will simply stop looking at the high level code as the end result works, in the same way most people stopped looking at generated assembly after 8 bit computers (for example, as most games were written in raw assembly for... perforamance), as it was good enough.
> If instead I use an LLM to rewrite a feature of a codebase I can't be sure that it still functions like the original one.
Right now, with existing static analysis tooling, you can ask it to write a full suite of unit tests capturing existing behaviour without modifying the existing code with 100% code coverage, and start there. Plus fuzz tests as well. I actually have marginally more confidence in that than a human being doing it.
It was difficult to not notice them. Opus 5 was unusable, most of my team went back to Opus 4.6 for most of their work. I hope we can move forward now.
How? Explicit instructions, memories and even skills have not been able to keep Claude from saying "genuinely" every two sentences and keep it from explaining heavily what something _isn't_.
It's also worth pointing out that the developers at Mozilla literally decided to make a brand new programming language (Rust) that eliminated an entire categories of bugs and started working on a new browser (Servo) where its components could be switched into firefox.
Then Mozilla fired the team working on it.
Lesson 1: Better tools and processes eliminate bugs.
Lesson 2: If you don't have the tools, make them.
Lesson 3: Mismanagement leads to security vulnerabilities.
Yes, and even after introducing and using Rust, there were still vulns in the remaining code.
FWIW, Rust is my main language, so I am grateful to the humans that created it, and it is indeed a monumental effort. But vuln finding is a different kind of creative endeavour, and one that is heavily weighted against defenders.
You would choose rust (from a security standpoint) because it massively reduces memory safety issues (which from a Microsoft study was responsible for around 70% of bugs).
Then you look at things like linting, fuzzing, design by contract, proofing, deterministic test runs, etc...
Yes, I should have been clear - I don't mind AI working on patches or vuln-hunting; my issue with MacOS 27 is the amount of AI included in the OS that is not opt-in. I use Claude Code everyday - but it doesn't run locally on my main production machines.
If Apple wanted to make the OS more secure, they would make the OS more modular, allowing users to reduce the surface risk by not installing bloated components
You have the same long list for Sequoia 15.8, so how is that better? Besides, newer OSes are also more likely to have uniquely knew bugs from both AI slop and in general all the shiny new broken features
I am currently on Sonoma and I was considering to upgrade to get the things that I missed in these years. I think and hope that macOS 27 will be faster wrt Sonoma...
Usually I wait until december to install the new iOS on my phone, since most of the time I don't care about the upgrades. This time I installed iOS 27 beta on my phone since the public beta was out, and I had 0 new bugs compared to iOS 26, but it hasn't been the big relief that I hoped for.
Now I am worried about what the new versions of iOS will bring us, I hope that the "we removed the crap" wasn't just a thing of this year, and that iOS will go back to high software standards. Phone prices are really high for me now, and I would like for my poor little iphone 13 to survive another couple of years.
So what's your experience in terms of speed and battery life with iOS 27 on the iPhone 13 so far? I have a 13 mini, and I still haven't gotten over the (perceived) shorter battery life with iOS 26.
Upgrading the battery on my iPhone 12 Pro has been a fucking revelation. I just don't run out of battery any more. I'm sure Apple was telling me it was only 20% degraded or something too, but I would strongly, strongly recommend this if thinking about battery life is currently an issue for you.
That could be a factor for sure, but this iOS upgrade basically broke mu iPhone as switching between apps got noticeably slower and sometimes the UI gets stuck for a second or two especially when switching between the camera and other apps. I'm so sorry I gave in and installed it.
I would say bettery is about the same of iOS 26, but mine is quite cooked (78% battery health) and I have to replace it.
Performance is for sure better. Not as much as I expected (I expected to notice it everyday), and I can't give you a number, but I remember that I noticed stuff being faster the first couple of days that I installed the beta, but now I got used to it
I reviewed the battery test results for iOS versions 26 (26.2, 26.3, 26.4, 26.6) and 27.
The comparison covered the iPhone 14 series and newer models.
The batteries in each model retain their charge 10–15% less effectively on iOS 27.
I concur with gekoxyz - I've been using the iOS and macOS 27 developer betas since about beta 3 on my personal devices, and had actually forgotten to a large extent that I was running beta software. It's been excellent - a marked difference to the 26 releases.
No, not exaggerating at all, but I'm not a power user of the phone. I just do really simple stuff, and I didn't find new bugs, just old ones. But obviously that's my experience... The most noticeable stuff was the app launch speed.
Interesting. Bugs on mine were from using core apps, IIRC. I wonder if bugs manifest on some iPhone hardware versions? Or maybe even which iOS version the phone was upgrade from?
It depends on what your definition of "small" is. If you consider the iPhone 13 mini as small the only thing that comes to mind is the Unihertz Jelly Max.
I am currently using both an apple watch SE 2 and a garmin vivoactive 5 simultaneously. Even if the Garmin has better battery life, I don't really like it for 2 reasons:
- the round watch UI is ugly and doesn't work with the other UI elements. Also I don't mind using a squared watch, I like it. My Casio is also square-ish.
- the UX is terrible, the customization is terrible. The default watch faces are all quite ugly, and the Garmin app makes the customization experience even worse.
Also, I feel like the sleep is more accurate on the Apple watch than on the Garmin from a couple episodes that happened and in general from the differences in tracking time.
Also, as you already said, I just charge the watch everyday while I shower, so battery life is not really an issue for 80% of the time. For the other 20% it's really annoying... but I think a newer model (and I hope that the new ones will be even better on that side) will have better battery life. When the battery life will be ~48 hours it will be perfect for me.
I am actively using Gemini flash to "translate" what Opus says into human language. I let opus do the design (with my assistance) and implementation, but then the report that Opus writes gets translated by Gemini so that I don't have to waste time to understand it.
> Wait — I don't understand where you've got to here. Re-pitch that: give me a little bit of context, talk in ASD-STE100 Simplified Technical English, and use the ubiquitous language from CONTEXT.md.
reply