having your own LLC should firmly place you on the side of "contractor"

One might hope, but I'm not a big believer in trusting the government to do the right thing. So they may "fix" the Uber/DoorDash problems and hose over a bunch of singles person LLCs. I know I can opt out of workers comp on the federal level, and I think I can also do so at the state level, but one of my clients has a dumbass "we require the businesses we work with to have X, Y, Z" lists with no process for exceptions and item Y is workers comp. So I pay like $7 a month for a workers comp plan for my zero workers (the owner isn't covered). All for the privilege of having this particular company operate as a middleman and take a portion of my fee. I'm actually doing most of my work for a different client that is more of a "we trust you to have whatever insurances you need, obviously if you set up an LLC you probably pay someone to advise you on all the important stuff, so carry on".

I'm not saying I know they will hose me over, just that I think it is entirely possible that it might happen.

Now the LLC I own is going to be forced to pay the only contractor (myself) more. Which is kind of a pain because what I'm holding back is my estimate of taxes and business expenses. I wonder if I'll be forced to buy healthcare via the LLC rather than directly.

The vision pro will have to deal with software written for the vision pro only.

It lets you "project" a Mac display into it (yeah, it won't run the MacOS software). It will actually tun iPad apps directly (yeah, ok, maybe of limited value, but it is a lot more then "just what is made specifically for the brand new limited market sized product").

Come up with an actual solution to an actual problem. This is the cart before the horse

I have one, but it is pretty niche. I do laser cutting as a hobby. My most recent laser cutter is a 40W diode laser in the visible spectrum. I have it in an enclosure with some class two laser safety plexiglass, which if the laser strikes it directly it'll just cut through. Unfortunately even catching the reflected light from the class three laser might be too much for it to prevent enough laser energy from escaping to harm my eyes if I'm looking at it (through the safety cage) unless I also have laser goggles on.

It is already hard to see what is going on without the goggles. With the goggles plus the plexiglass I can't see squat, so I can start my burn and get no real feedback until it is done.

With the vision pro on my face I could just open the enclosure (and disable the safe interlock) and look in. I mean, yeah, this is a real niche because it requires someone to spend $3k to be able to get a "good view" of what is going on inside their $2k laser cutter. So that is a world wide market of, maybe just me, or maybe there are like a dozen of us.

I thought I saw it having a battery pack hanging off it....is that not hot swappable?

I don't know if it is hot swappable, but it is apparently a USB-C battery, so you could use a larger capacity battery rather than the one they ship. Or even a USB-C power supply as opposed to a battery.

That said I would hope it is hot swappable.

Why do they have to even have an idea of what the missing application is? The point is to create a canvas for all the app developers, isn't it? They are banking that someone comes up with the "killer app" using their tool. Sometimes you don't even know what you can make unless you have the tool first and can experiment with it.

With the iPhone they showed the "missing application" when they announced it. With the AppleWatch they showed stuff they thought was missing, and stumbled into the real missing app after shipping.

If Apple knows what the killer app is before they launch the device, they will attempt to ship it. If they never figure out what the killer app is the device isn't going to overcome its immense price tag. Which is significant because that means even if they somehow hit a killer app for you they never reach market levels that encourage them to stay with the product line. They also never reach market levels that will cause 3rd parties to rush to the platforms to try to get $1 from some significant fraction of everyone there, because that'll be like $37 and not worth the cost of buying dev unit or spending time on an app.

In other words as anyone except an interested observer there are real reasons to hope Apple figured out what the killer app was and addressed it...or at least ot hope they do figure it out and manage to address it.

To some extent it doesn't matter if Apple figures it out or if someone else does...except Apple has had this for quite some time now, if they haven't figure it out maybe nobody else will either. Also Apple gets to shape the hardware, so if the killer app is something that is in sight but out of reach they had a lot of time to figure that out and adjust the hardware so it whatever it is that the VisionPro mad e(almost) possible was something they could actually do. 3rd parties have to be content with the hardware as shipped. Likewise Apple has some advantage in terms of software APIs and OS capabilities. If the killer app requires integration between multiple things Apple can do it, but likely that siloed sandboxed 3rd party apps can not.

this place closed in less than two months. I don't know if it'd be correct to label it as part of the expected failure rate

Yep, anyone that takes advice when opening a restaurant will be able to run it for six months to a year without any actual income. Except that many many places underestimate how much it costs to get a place open in the first place, like all the months you are paying rent on a commercial space waiting for health&other inspections before you can open.

So if they followed common practice and set aside enough money and still closed in under two months it could "just" be that it took way longer for them to get open then they had anticipated and no longer had months and months of runway to get to profitability. It could be that they didn't follow common practice and thought somehow they would open and instantly be able to more then cover costs, and in reality they like most places don't start out as popular enough to stay afloat.

It is also entirely possible that they didn't know anything about running a restaurant, or business in general, and didn't seek out advice of people that had already done so, or that they did, but decided they knew better and based on a spreadsheet that says they can make smoothie in 2 minutes and opening from 10am to 4pm is 180 smoothies per day per employee, set the per smoothie price to make that all work out...and then were shocked to see that being able to make a smoothie in two minutes only matters if someone orders one...

assuming that actual AI was involved and they weren't merely stretching the meaning of the term for their own small shell script of an order-taking system

I assume it made bespoke recipes, maybe something per customer. I also assume they were not in reality any better then non-AI generated recipes, but if customers think they are and visit you for their customer smoothie as opposed to Jamba Juice for the Mango-a-go-go it is a win for your AI powered business. Hopefully that doesn't encourage them to get into a paperclip making sideline...

If you try and mitigate this by turning it down a bit, you just get copies of existing smoothies. Seriously, look up articles on what happens when you ask an electronic parrot to create a recipe.

I use to work at search/ad company that had an AI sideline. Several of the cafes served items from AI generated recipes. The only ones I ever tried were cookies. They were fine. They did have a few ingredients that I wouldn't have expected like chyane peppers in a chocolate chip cookie. I talked to some of the kitchen staff about it, I was wondering if they just skipped dumb looking ingredients, or if more cookies had peppers in them then I had expected or what. They did the recipes as given because it was part of an experiment and while the kitchen staff isn't particularly technical they like to support their local tech staff. Plus they thought the peppers were weird ingredients and wondered if it would make any difference. Cooks like to experiment too, and normally they don't get to just whip up 800 lbs of cookie dough with a weird ingredient on a whim.

So I asked how many AI recipes had weird stuff in them, and the answer was lots. Basically they end up with something in them that doesn't belong, and it either has a lot, and people say "this cookie sucks" and that ingredient gets substantially reduced, or an ingredient starts out with "just a smidge" and that recipe doesn't do better then any others so it doesn't significantly vary the "smidge". So the AI recipes accumulate trace amounts of things that don't work because the cross pollination step of that particular AI system doesn't try to zero out near zero items just in case they turn out to be important later. It is useful in some contexts. Somehow I doubt the 800 lbs cookie recipe with one ounce of cooked beef is ever going to usefully mutate to one with enough beef to taste (but: maybe it is a useful fat contributing to the texture of the cookie? Gah! I'm why programmers shouldn't design systems to make recipes! Fine! Ok, I mean realistically in an actual recipe the cost of each ingredient is important, as is the labor of managing "one more thing" as is the time taken to measure it and combine, so really if that ounce of beef isn't contributing anything keeping it "just in case" it is useful in a future variant isn't useful, and using it rather then slightly more butter or whatever oil is already in the recipe is also a downside...then again zeroing near-zero ingredients is likely to make you more susceptible to the hill climbing problem. Beh!)

So 8GB of RAM is now meager? Not long ago it was plenty because ARM used so much less RAM than Intel!

CPU Architecture has a minimal impact on memory usage (at least if using the same bit width for pointers and integers!). On modern CPUs the amount of RAM occupied by instructions is pretty small compared to image assets, sound samples, user data and the like.

That said x86-64 has a pretty compact instruction stream, and likely is more compact then traditional ARM instructions which are all 32bit wide and âoewasteâ 4 of them on making every instruction potentially conditional. The thumb encoding is likely to be slight more compact then x86-64 although it shares some of the common drawbacks of CISCs despite not being one, not boring able to use all GP registers for any given instruction, and only having two encoded operands sometimes you need an extra instruction to copy a value to another register.

I have seen a lot of people claim the M1/M2 donâ(TM)t need as much RAM because âoeunified memoryâ is magic, but that too is BS, at least compared to systems with dedicated VRAM, it has some upsides, but reducing total RAM usage really isnâ(TM)t one.

The most likely two reasons people think M1/M2 systems donâ(TM)t need as much RAM are (1) wishful thinking, and (2) the SSD subsystem on those machines is insanely fast so paging doesnâ(TM)t hurt as much. It is still far far worse then having enough RAM to not page, but at say half the penalty being a little shy on RAM hurts less. Like a bruised shin rather than a sprain. I would rather avoid both, but one lets me limp around saying âoeIâ(TM)m good guys, Iâ(TM)m good, letâ(TM)s go!â

I didn't mention m4 because I think it is a big deal, or that you can't use it elsewhere, it was an attempt to show "I actually know something about use of X11, it isn't something I head of but know it is old so I'll pile hate on...or even I used it and didn't like it so I'll pile some hate on", it is something I used at a deep level, made code for, and enjoyed.

I know X11 has some alpha compositing extensions, I didn't know they really made it into widespread use.

I also wasn't really attempting to compare X11 to Wayland because I have basically zero experience with that (I have some RPi systems that can be configured to use it, but I use all my RPi systems of headless tasks so I haven't attempted to use X11 or Weyland on them!)

I wouldn't say OSX is uniformly superior, but it has some a vast number of areas it is head and shoulders above. I personally don't like it's window management UX as much, but most people do actually like it more (and to be honest if I went back to X11 I bet I wouldn't enjoy fvwm as much as I once did, although I think I would still like point to type more then click to type). I definitely like the idea of window managers being something the user can pick one of as opposed to it being bundled with the OS and if you don't like it too bad, find another OS or learn to like it. Pretty much any other part of X11 though is inferior to macOS and iOS graphics systems. Oh! Except maybe that macOS/iOS have too many choices on how to do almost the same thing. Like UIImage/NSImage can be backed by a CGImage or a CIImage. Why? Well CGImage is a pixel array of some sort while CIImage is some sort of recipe for taking filters and parameters and frequently pixel images (or sometimes devices) and turning them in to an image. Why are they different? Why couldn't the be unified under one API? I mean ok, it isn't the end of the world, but it is an odd little corner that is sometimes quite irritating.

I share multiple accounts with my husband. How are they going to allow us to both access them from different locations at the same time? Turning off 2FA and using a long password that we both can remember is the best we can do for now - unless they allow more than one passkey to work on the same username this is a show stopper for us.

On iOS/macOS passkeys are managed by the same system that manages passwords, and you can manually share them via AirDrop. Once shared two accounts can use the same passkey. In iOS 17 you may be able to place passkeys in a shared "iCloud container" to make that easier (I don't know if you can because my wife hasn't upgraded yet so I haven't tried the cross account iCloud sharing).

There is no reason other password managers can't do the same type of thing, and in fact other password managers have supported cross account sharing far longer then Apple has.

There are no games on MacOS

Well there is WoW, but that is basically the whole list of games that are released on the Mac at about the same time as for Windows. It has a few other items that come out "later". The only way gaming will ever be a thing on MacOS is for Apple to invest in it, just like they did for the iOS App store

From time to time Apple makes a push. Most recently they did a whole "game porting toolkit". Apple's issue here is less the technology, but more the lack of both a multiyear commitment, and lack of desire to form relationships with game companies and treat them as peers or at least relationships to manage well. The don't like to receive input and give roadmaps. They like to once a year announce what they have released and have others rush to adopt. It works well when they are the leader in an area, and poorly when they are a distant third or fifth place. The app store wasn't a success because it was technically interesting. It was a success because the iPhone was a huge market and the app store was the best way to reach it (mobile safari web apps were also "a way" to reach it, but a distinctly second rate way).

It is not perfect but it's a pretty good protocol/API.

It is kind of awful. Or at least significantly subpar. I use to work on twm, tvtwm, and fvwm, or an xtrek, nettrek, and xtank. Not like "I used those to do work", but I contributed code to them (or was a primary maintainer). For example it is "my fault" that many X11 window managers use (or have modes to use) the m4 macro preprocessor on their config files. I won't say I'm a world expert in X11, but I have done more then my fair share of work in and around X11.

X11 was an ok, but not awesome design to support what was common in the early 1990s. It didn't keep up with changing hardware or users needs/expectations. Font handling for example was poked at a few times and then largely left to rot. I would take MacOS's or iOS's graphics subsystem any way of the week and twice on Sunday for most jobs. Rendering fonts and curves is far simpler, anything dealign with alpha transparency is easier, and so on.

For a recent project I have been controlling a laser cutter. I ended up using macOS's graphics layer to render fonts as BezierCurves and combining them with other BezierCurves (most of macOS's rendering layer deals in BezierCurves, and you can get it to cough them up in many contexts), and I collect the curves into collection of CutPaths withe various nongraphical attributes (laser frequency, power level, and number of passes). I form them into tiles, space the tiles on a surface (the tiles are "mostly" rectangular, but I have some support for fitting non-rectangles when it is really needed to pack things onto a single sheet of material). I convert the Bezier curves into things the laser turret handles and burn my projects. X11 would be utterly unsuited to the task. Now to be fair most of what I get from macOS isn't hard to do directly with my own BezierPath code, making curves out of straight line rectangles and line segments is easy enough, ellipses are harder then you would think, but with an assist mathematica it is a solvable problem. Doing text layout is a stone cold bitch though. Before deciding to lean on macOS for bezier paths I took WebFotns which are a collection of SVGs which aren't too hard to turn into Bezier Curves. However even after writing a significant amount of code to parse web fonts, and then lay them out character after character they look awful in some combinations. It turns out kerning and ligatures were invented for a reason, and the reason is because text kind of sucks without it. That might be acceptable in some contexts, but if you are layering text onto things it mostly isn't a debugging "SLOT A" it tends to be the important decorative embossed text. X11 might actually do some of it, but not in a way that lets you extract anything other than pixels. Not a curve. Plus borrowing macOS's text system I can do things like "lay this text out across this other curve not straight lines. Or "lay this text out in this rectangle, but exclude these other sub areas in the rectangle..."; X11 doesn't do any of that.

Those seem like minor things, and to a certain extent they are minor. On the other hand basically anything you look at in X11 has a similar deficit. Except being able to shove graphics operations over a network connection. Other windowing systems basically go pixel diffing of virtual frambuffers and send compressed diffs along. If that doesn't work well for a particular task you are screwed. X11 shoves graphic primitives over the wire, this is much better for some applications. You can also do the virtual frame buffer diffing thing with X11 if that works better for your application.

X11 could have used an X12, but never got one. It should really be about X17 by now. That doesn't mean it is worse then Weyland, but trying to defined it like it is an awesome piece of software as opposed to something old and creaky that kind of gets by isn't really helping address the issues X11 has, and by extension the drag it has on everything built on top of it.

How the hell do ya'll haul stuff home from the store on your bicycles???

I don't. I use to work at a place that had bikes to get between buildings on a large campus, and I didn't like taking a laptop in the bike. In the basket it just seemed like it would hop out and get wrecked, and in a backpack it made my balance all lousy...

However they do sell cargo bikes, basically bikes with a huge almost person sized basket, or more or less trailer that hitches behind the bike. I had a eScooter with a "cargo pod" for a while. It was pretty usable.

Now I live somewhere with winter snow, and all that stuff is pretty useless with a few inches of snow on the roads.