CharlesW

This is technical content, written for a technical audience, being shared on HN. In that context, it's not particularly esoteric. If "5x faster Edge Functions" means nothing to you, that's a good signal that you can just skip it without FOMO. If you're familiar with the concept but not Firecracker (a popular microVM) specifically, the expectation is that you'll use search engines and/or LLMs to get up to speed.

> This created things like roentgenizdat or "ribs" (gramophone recordings made from polyester X-ray films)

This is a plot point in the excellent Star City, a spin-off of For All Mankind. https://en.wikipedia.org/wiki/Star_City_(TV_series). Also see this project by English musician Stephen Coates: https://www.x-rayaudio.com/

From a systems perspective, the brown M&Ms are a "canary" (a.k.a. "sentinel indicator"). If a venue can't handle that or simply decides that it's "mega baloney" and ignores it, one can no longer assume that they've done anything competently, or at all.

HN mods [*generally] don't magically appear here when mentioned (unless you do it three times in a mirror while sacrificing DRAM), but they are absurdly responsive when contacted via hn@ycombinator.com.

> I find myself forgetting the point. The goal might not be to exactly enjoy every sandwich. Some sandwiches are just straight up trash. Maybe the goal is to notice that you’re eating one.

Taking Mr. Zevon's advice to "enjoy every sandwich" literally is amazing. I think he'd appreciate the absurdity of the post.

I wonder if the post's focus on Linux/Android is just because that's where they started, or because they're leveraging techniques that are only possible on Linux/Android?

> Since it never got past internal demo (and use in PPC in some ways)…

For anyone who might not know, parts of it did forcefully and successfully escape containment. Core Mac Toolbox APIs originally created for Star Trek (like the portable C implementation of QuickDraw 2D) were installed on hundreds of millions of PCs as QTML, which QuickTime for Windows was built on, which was in turn a requirement for iTunes for Windows. QTML on Rhapsody was also the PoC for Carbon for Mac OS X, with ClarisWorks as the guinea pig.

https://archive.computerhistory.org/resources/access/text/20... (see page 28)

Apologies, I didn't mean to upset you. Per the title I was talking about the downloaded core model sizes, which are a bit under 8 GB or 14 GB in OS 27 depending on device: https://support.apple.com/en-us/121115

Yes, I'd sanity-checked the answer on my machine, which is 15.5 GB on iOS 27.2: https://imgur.com/a/XVT8rme

Not all of the space used by Apple Intelligence is downloaded model size, which explains the reasonable difference.

This is a good overview of how this is done: https://www.anthropic.com/engineering/demystifying-evals-for...

Interestingly, Siri AI was able to answer this with a reference: Up to 14 GB. https://imgur.com/a/SZGh1ca

I appreciate the detailed answer, thanks!

> it's relatively useless (even with Siri stuff they added in macOS 27)

Would Antfly do better with the tasks listed at https://pogueman.substack.com/p/125-tests-of-the-new-ai-siri?

> Tongue half in cheek, my point is that agents may be unreliable with fresh libraries.

You may find Axiom (my FOSS project, updated for iPhone Duo) helpful. https://charleswiltgen.github.io/Axiom/

Note: I'm scrubbing it against the final 27.1 beta SDK as I type this, so it's likely that its Duo knowledge will get an update within the next few hours.

I don't understand what you mean by "eventually" when you can't even pre-order it yet.

It's mostly enforced, the only exception being apps with a legitimate iPhone-only hardware dependency.

iPhone apps must function on iPad in iPhone-compatibility mode, apps must support window resizability as of OS 27, and apps can't detect iPad and refuse to run. And I don't think this is new, but setting the target to iPhone-only doesn't prevent iPhone apps from running on iPad.

Correct.

> Also, the amount of work to support this with zero support from Apple since WWDC is just absurd.

Resources:

https://developer.apple.com/search/?q=iphone%20duo

Axiom: https://charleswiltgen.github.io/Axiom/ (my FOSS project, updated for iPhone Duo)

> It is indeed a niche product.

It's worth noting that "niche" in this case means ~10M in its first year, comparted to an estimated 250M Apple smartphone shipments in 2027. So far, Apple has sold ~600K Vision Pro in total.

Also, supporting Duo is not an enormous lift if you're already supporting resizable windows, as developers are required to do for OS 27.

Fair enough, knowing when to be and not be prescriptive about implementation details is critical too!

> As effective as “make no mistakes.”

Those aren't comparable instructions. Providing useful, related context to improve outcomes is a basic best practice, and asking LLMs to do research first is an important source of that.

That’s “brown note”:

https://southpark.wiki.gg/wiki/The_Brown_Note

https://en.wikipedia.org/wiki/Brown_note

Because they're experimental. Since that's confusing even with a key, it may be better to hide experimental features/categories by default.

> We can completely leave to the side that your weighting scheme is surely arbitrary or perhaps even biased.

Hover on any number to see both "raw" and "with experimental/non-standard features" scores. If you do this to the overall score, you'll see that Firefox benefits the most from this.

The math is correct, see /about for methodology and this summary from the "How Scores Work" panel.

   About PWA Scores:

   • The main score shown is weighted for feature importance
   • Only stable (non-experimental) features are counted
   • Tap or hover any score to see:
      – Raw scores (simple % of supported features)
      – Experimental feature scores

> edit: also, your site is just obviously biased/poorly done

If you can offer recommendations for data-based improvements, I'd sincerely appreciate it! It was made without any concern for Safari "winning", as is probably obvious since it has the lowest score. Methodology described here: https://pwascore.com/about

It's not ad hominem since I'm critiquing their position, not their character. Their positions are public and well-documented, which is why it's important to point out the bias of what you linked to.

> And, no, safari is NOT capable of PWAs…

Incorrect, see my link. PWA installation on iOS is simple and idiomatic, and the same as it is for all websites: Share > Add to Home Screen.

Note that Alex Russell is a Blink partisan, currently supporting Blink at Microsoft and formerly having done so at Google. Folks like that don't take kindly to the idea that a Blink monoculture might not be a great thing, and definitely do not share the opinion that it's critical to the open web that Apple not cede the open web to Blink.

Remember, Blink began as Google's 2013 fork of WebKit. They've embraced and extended it, but thankfully the "extinguish" step is taking longer than they thought. This makes the "open web, as long as it's my engine" people upset.

Safari's PWA support is not as complete as Chrome's today, but it's as good as Firefox's support and perfectly capable of supporting rich, standards-based apps. https://pwascore.com/

Yes, it appears the submitter rewrote the title (strike 1) without even reading TFA (strike 2). Not great.

Actual title: "New MIT Algorithm Meets Every Hard Constraint in Simulated Tests"

> It'll be interesting to see how many/how quickly apps adopt the UI that lets you make the most of the foldable screen.

For iOS 27, apps are automatically automatically opted in to resizability. Apple was clear at WWDC that apps need to dynamically adjust to any available scene size at runtime, not just for iPhone Duo, but also for iPad windowing and iPhone Mirroring on Mac. There may be a grace period, but Apple hasn't shared one AFAIK.