You win btw.
1,346 karma · joined August 3, 2021
This problem is not unique to mapping objects to SQL databases, but mapping objects to anything remote at all, say a GraphQL or a REST API.
OOP as we presently interpret it is an inherently synchronous, reference (or handle, rather) based paradigm, that only works when all your state is local and in RAM.
This is why distributed object protocols keep failing. They'll keep failing until OOP programming reorients to use values for messages and makes references explicit, so their impact is seen and felt (and it's especially seen and felt when you reference an entity on another machine halfway around the world, in terms of lag, fragility, eventual consistency and everything).
We see hints of this with value types in Swift and .NET, unfortunately I'd say rather rudimentary so far. But it's coming. The ideal example of such a set up is Erlang. A language that Joe Armstrong has called "probably the only OOP language in the world". A statement Alan Kay agrees with (source: https://www.quora.com/What-does-Alan-Kay-think-about-Joe-Arm... ).
And I bet Gmail.com running in a modern browser is more complicated than a mail client from that time.
As such, browsers are now an app platform. They don't need to ship with prebaked apps like mail. They need APIs like a powerful runtime, visualization layer, background services and notifications... and they have them.
In a way you describe how we can emulate this process. The question is why the heck wouldn't databases do this themselves? Same with adding and dropping columns.
Consider how PostgreSQL encodes null for example, by skipping them in the row as fields, and adding them in a null bitmap in front of the row. Meaning... rows are not uniformly sized, there's no math like offset = row * rowsize + field_offset; kind of addressing for reading a field in PG where recoding some of the rows breaks the entire table.
And yet we have all those huge monolithic operations that need to be done atomically. So weird.
You see, the rules of formal languages that encode formal rules of system constraints pre-date computers by centuries. Think of math proofs, for example. Sure, we can encode symbols as emojis, or geometric figures or whatever. But in the end, it's sequences of symbols, that's the nature of it. And tapping symbols one by one with a headset will suck, no matter how programming looks.
The rules of formal languages that encode formal rules of system constraints pre-date in fact our species too. Think about what DNA is. Oh yeah, spooky, isn't it. A sequence of symbols (GTCA) encoding a sequence of more complex symbols (proteins). Spooky! But yes, DNA is our code. And it works the same as our programming code.
Now I know where you're going. LLMs. Let's assume an LLM writes the code for you. You still have to read it, which you can do fine with a headset (if it's not as encapsulating and heavy, and with short battery life as Vision Pro v1). But if you spot something's off, you need to adjust it. Go directly for the kill, and make that surgical series of edits. You know? Or... maybe you can spend the rest of the day hopelessly trying to explain to Siri 2030 year edition what you want to do, instead of going in and doing it, for that "last mile".
Because if AI can do the last mile itself, to the point you don't need to even verify it... first, that's the fast way to AI shipping code we don't understand and basically giving up our entire civilization to it. And second... we don't need to code, but we also won't need to exist, and therefore not need headsets.
So in the worldlines where we DO exist... Vision Pro sucks for coding, because it's a shitty human interface to editing code.
And in the worldlines where we DO NOT exist... Vision Pro sucks for coding, because AI doesn't need headsets.
I'd expect a similar wave of people breaking their expensive Vision Pros if they try cooking with one. It's a terrible idea. First, most kitchens are cramped, full of low-hanging cabinets to break your Vision's fragile glass into.
And then, keeping those open vents around vapors full of fat and tasty food bits is a great way to cover the circuits with grease.
We already have a solution for something telling you what to do next, and it's called an iPad with a stand. A phone also does the job and has much lower chance of incidents than a headset, despite yes, you may need to wash your hands from time to time to scroll down. Or... you can simply use assistive features and voice for that. Siri is going to get a lot smarter thanks to LLM, much sooner than Vision Pro will become light and pragmatic for such purposes.
Regarding this "it'll know where things are in your home", let's use basic logic here. It can learn the layout of your rooms and where your immovable furniture is. But no, it can't know where everything that moves is, because this means you literally can't move anything unless you have the headset on to track its location. Or slap expensive AirTags on every single jar and utensil maybe. All solutions would be hilariously impractical. And... we'll end up with where I started: a broken Vision Pro glass as you slam it in a cupboard while trying to fish out a jar of condiments.
I don't know what is about VR that makes people pull out the fantasy scenarios. It's simply a (bulky) screen with pass through. It's not a wizard. It can't know things unless there's a way for it to find them.
We can imagine a super-thin model that you can keep on your face 24/7 and sleep with it too, so it tracks your entire life forever and knows you better than you know yourself. And it synchronizes with your spouse and children who also wear their own headsets 24/7. And it's unbreakable. And the battery never runs out. We can imagine many things. They don't exist, and won't exist any time soon. "Not on the horizon" as Steve Jobs used to say.
The funny thing about this statement is that Vision Pro is in fact the smallest rectangular screen of any device he's written code for.
He's hyped up. That's normal. In time he'll understand Vision Pro doesn't provide any better UX for common activities. In fact it's worse in many ways.
Where Vision Pro may shine is tasks where you need to perceive and manipulate complex three-dimensional objects, as they would be in physical space. I see great uses in engineering, design, art. It'd be great to preview interior design, design cars, architecture, create machinery and so on.
It'll also be great for previewing products, so online stores become a lot more viable than they are now, as you get a sense of size and style for an item in Vision Pro.
It may also be great for education, training, simulations.
It has many great uses. But basic apps isn't it. And most people won't care. This thing sucks to wear for more than 20 minutes. It's heavy and uncomfortable. You can't share your experience with others, either. It costs a lot. And you can't multitask with it. I can walk to a place and do something on my phone.
The input model also sucks. To code, for example, you need to hook a bluetooth keyboard and mouse. Looking at symbols one by one to fingertap would be comically slow. At which point, you may as well just get 2-3 screens and work on a normal workstation. For less money.
Also is it a fact that people of certain origin are overrepresented in given populations, classes and so on. The solution is NOT to turn a blind eye to this, or lobotomize the AI in order to conceal it.
Also do you realize he paid about 1/4 of the value of Twitter. The rest is loans and other investors burning cash by trusting him and rolling over.
Tell me why did a wealthy man who merely spent "20% of his savings" take on high-interest debts worth over 13 billion?
Clearly Elon doesn't even have the decency to file for bankruptcy. He'll following the Trump book. Bold claims, bouncing checks, blaming everyone else and playing the victim.
(Like rain on a wedding day)
It all hinges on details of their contract (or lack thereof), and Google has to tread carefully because the black PR of Twitter dragging them around as a result (and you know very well Elon is GREAT at playing the victim) will result in massive losses from other businesses using Google Cloud.
1. Elon introduces Twitter Blue $8 subscriptions and announces that when enough Blue subscribers block someone, the victim will be rank demoted for the entire site.
2. Twitter Blue rightwing subscribers start organizing block mobs attacking the left and generally people they don't like, hoping to in effect force Twitter to shadowban "enemy" accounts. It's working.
3. This forces potential victims to block Blue checks on sight, immediately, because otherwise they risk being blocked back, and rank demoted.
4. Elon complaining publicly that increased blocking activity is like DDoS on the timeline service.
5. Elon saying how blocking is not nice and he has removed all his blocks on his account, trying to influence his followers.
6. Elon testing the waters in this tweet that the block feature is going away, to see if there will be a backlash.
He wants to remove it. Everyone is blocking everyone due to the climate of Twitter these days, and all these poorly thought out ranking features Elon has forced on the platform.
I've also missed part of the story where at one point there were browser extensions "block the Blue" that were mass blocking anyone with a paid check. Elon banned accounts promoting such extensions directly. He was terrified of the aftermath on the social graph.
The timeline service now has to work twice as hard implementing these giant blocking lists. So now he wants blocking GONE.
Different people give you instructions. But you never see them, only the paper with text on it. Some people are "admin" some are "user". You have to guess from context. But instead of identifying themselves, they're incredibly vague about it all, at best you may get something like "User:" before a line, to tell it apart from the ambient instructions you were given beforehand.
And somehow it's your fault if you misidentify who is supposedly writing some part of the text.
This is not "gullible", it's poor signal for the model. It has no way to know who is who, it's all the same token stream to it. No voices, no faces, no caller id, nothing for it to hang onto for recognition. What is it supposed to do?