I think the author makes a hard distinction between consumer products and infrastructure/engineering products. The Shelby Cobra has a funny name, but its engine is the memorably named V8. The Hoover Dam is a dam, and the Golden Gate Bridge is a bridge.
We can argue about namespace pollution and overly long names, but I think there's a point there. When I look at other profession's jargon, I never have the impression they are catching Pokemon like programmers do.
Except for the ones with Latin and Greek names, but old mistakes die hard and they're not bragging about their intelligibility.
Yeah, V8 is the shape of the engine - 8 cylinders in two rows offset at an acute angle (i. e. V-shaped). Likewise a V6 has the same number of cylinders as an inline 6 but performs very differently. There's a handful of different engine shapes - I'm fond of the rotary engines used in early aircraft. Traditionally, the name of an engine was just the year, the manufacturer, and the displacement (like 1965 Ford 352). You often leave off the year and even the manufacturer if it's not required by context.
The Ford 351 is a bit special because there were two different engines made by Ford in the same time period with the same displacement, so they tacked on the city they were manufactured in (Windsor or Cleveland).
That would be a laudable goal, but I feel like it's contradicted by the text:
> Even on a low-quality image, GPT‑5.2 identifies the main regions and places boxes that roughly match the true locations of each component
I would not consider it to have "identified the main regions" or to have "roughly matched the true locations" when ~1/3 of the boxes have incorrect labels. The remark "even on a low-quality image" is not helping either.
Edit: credit where credit is due, the recently-added disclaimer is nice:
> Both models make clear mistakes, but GPT‑5.2 shows better comprehension of the image.
Yeah, what it's calling RAM slots is the CMOS battery. What it's calling the PCIE slot is the interior side of the DB-9 connector. RAM slots and PCIE slots are not even visible in the image.
It just overlaid a typical ATX pattern across the motherboard-like parts of the image, even if that's not really what the image is showing. I don't think it's worthwhile to consider this a 'local recognition failure', as if it just happened to mistake CMOS for RAM slots.
Imagine it as a markdown response:
# Why this is an ATX layout motherboard (Honest assessment, straight to the point, *NO* hallucinations)
1. *RAM* as you can clearly see, the RAM slots are to the right of the CPU, so it's obviously ATX
2. *PCIE* the clearly visible PCIE slots are right there at the bottom of the image, so this definitely cannot be anything except an ATX motherboard
3. ... etc more stuff that is supported only by force of preconception
--
It's just meta signaling gone off the rails. Something in their post-training pipeline is obviously vulnerable given how absolutely saturated with it their model outputs are.
Troubling that the behavior generalizes to image labeling, but not particularly surprising. This has been a visible problem at least since o1, and the lack of change tells me they do not have a real solution.
Eh, I'm no shill but their marketing copy isn't exactly the New York Times. They're given some license to respond to critical feedback in a manner that makes the statements more accurate without the same expectations of being objective journalism of record.
That's a very good technical solution, but socially it can be foiled by an official-looking alert saying "failed to scan card, please do X instead".
And that's assuming the technical solution is deployed everywhere. I'm in the EU with one of those IDs, and I still had to upload photos of my passport and scan my face to open a bank account. The identification process even had its own app that I had to install.
But then again, should the EU follow up with a similar policy, it could mandate the use of these checks and prevent/penalize ID photos. I’m very optimistic here.
Exactly. I'd concede this point if I'd seen a giant public awareness campaign informing people which official sites to use and general safety awareness about it. I can tell you, literally nothing like that has happened. Not an insufficient effort at it - no effort, nothing. It's clear the people in charge are just head in the sand about this aspect of it.
The most impressive "fuck off contact page" I've seen was from Trade Republic, an investment app. The support page has a QR code to the in-app FAQ and nothing else.
Turns out that a handful of FAQ answers have a chat widget (with a chatbot, of course) that can be coaxed into switching out to a human. But if your topic is not on the FAQ, the answer doesn't have a chat widget, or you don't randomly click around other topics, you'll never find a contact form.
Even the "complaints" email address found in their legally-mandated Impressum just auto-replies with instructions to use the app help. I've since closed my account, but I'm still amazed how a company holding people's money can shield itself so completely from customers.
Cool project, I wish we had more GUIs for these OS functions. How was your experience with GTK4 and Rust?
And it's a bit sad that in the year of our lord 2025, the best way to get such fundamental information is by using regexes to parse a table[1], generated by a 6000-line C program[2], which is verified by (I hope I'm wrong!) a tiny test suite[3]. OSQuery[4] is also pretty cool, but it builds upon this fragile stack.
That's something I miss from Windows, at least PowerShell has built-in commands that give you structured output.
> It's a bit sad that [we're parsing output] ... generated by a 6000-line C program[2] ...
> That's something I miss from Windows, at least PowerShell has built-in commands that give you structured output.
It sure is something to disparagingly point to the LoC of 'ss' in one sentence, then pine for both PowerShell and the Windows infrastructure that supports it in the next.
You mentioned processing the output with regexes. That's definitely a code smell, but this is one line of the data from the 'ss' command in question, with fancily-aligned header line included, but with vast tracts of whitespace removed. The regex you pointed out is processing the column whose comma-separated data is enclosed in parens:
Netid State Recv-Q Send-Q Local Address:Port Peer Address:Port Process
tcp LISTEN 0 666 [::]:22 [::]:* users:(("sshd",pid=1337,fd=7)) ino:1338 sk:2024 cgroup:/openrc.sshd v6only:1 <->
They definitely didn't have to use a regex to process that, but chose to.
You could argue that a system that let you write client code that goes something like
is superior to one that requires writing something that makes use of the moral equivalent of 'cut'. I'd argue two things, one of them informed by my professional experience with PowerShell
1) What happens when the "structured" data you rely on changes shape? When that system that produces that "structured" data changes 'users' to 'user_list', 'cmd' to 'local_command', or deletes 'process' and moves 'users' up into its place, you're just as screwed as if 'ss' changed its output format in a way that wasn't backwards-compatible.
2) The core Microsoft tools might all produce "structured" data, but -in my professional experience- so, very, very little "community-provided" PowerShell code does. Why? I don't know for sure, but probably because it's notably more difficult to make a script or library produce that sort of data than it is to just emit regularly-formatted text.
The problem of data changing shape can happen regardless, but with text you have the added danger of escaping characters and ambiguities. Not to mention there are ad-hoc text formats for each and every tool, which can change from one version to another.
And you're right, PowerShell is far from perfect. I miss some of its design goals, not the whole thing.
I've always found Retinal Waves[1] interesting. During development of the visual system, there are spontaneous bursts of activity without external stimuli, helping the synapses to organize properly.
In my layman's view, it's like hallucinating shapes that are important to learn. Very similar to the "priming" described in the article, but easier to visualize (literally).
I feel like the retina, & sight-brain connection more generally, will turn out to be a lot more important to human cognition & consciousness than we realize
AFAIK the whole morphology is decided in distributed computation fashion via electrical potential changes, at least according to experiments by Michael Levin
Very sleek marketing, but why did they rebrand (the fantastic) KDE Connect to "Zorin Connect"[1]? From the mere <30 commits, I see no reason for the fork, only confused users.
If it was tightly integrated into the OS I could sort of understand not mentioning its name, like you don't want "Foobar Control Panel" and "FizzBuzz Start Menu". But KDE Connect is a standalone app you can install even on Windows. And this is not just hiding the name, it's replacing it!
So, why the "rebrand"[2]? It feels like an attempt at stealing credit.
Presumably because a good portion of the target audience are less technical users, and there's no reason to be throwing extra (ex-)initialisms at them. "Zorin Connect" is clearly something that will connect my Zorin OS to something. "KDE Connect" is something that might perhaps connect me to a "KDE", but what's a "KDE" anyway and why would I ever need to connect to it?
So that’s how you’re getting dependent on them even more. So probably you need the pro version with the paid support.
Stealing credit was the definitive factor for me, I wanted to try it recently for replacing Windows for some folks, but now I think I’d just go with immutable Fedora.
Forking it makes it easier to convince your flock of sheep that you must pay for GPL software. It also gives them a lot of opportunity to inject their own happy accidents into there.
Growing up is also understanding that the opposite of bad isn’t good. In the same vein, Ms, Google, and Meta being bad, does not help Zorin OS being bad either. Red Hat is bad too, apparently, but I have no issues with using Fedora.
Step 1: force websites to add an opt-out flow for privacy-minded users.
Step 2: websites don't complain too much because they can implement it in obnoxious and dark-pattern-laden ways, so that few users actually opt-out.
Step 3: now that websites have proven there's no technical barrier and the flows are already implemented, slowly retire unnecessary user tracking and data sharing.
I'd be surprised if this was planned ahead of time, but it's not a bad strategy.
We can argue about namespace pollution and overly long names, but I think there's a point there. When I look at other profession's jargon, I never have the impression they are catching Pokemon like programmers do.
Except for the ones with Latin and Greek names, but old mistakes die hard and they're not bragging about their intelligibility.
reply