As you delve further into the article, and (ironically, based on the topic) consider the ideas presented, it becomes more clear. Or skip to near the end to see.
TLDR: AI and generic pop culture artifacts have something in common - a lack of a certain type of depth that makes art great.
I think that's not the author's point; he's mostly talking that time solidifies ideas as shown by the evolution of points of view. At the same time, his picture of the wave shows that over time the painting feels more distilled and simplified. So is his point actually that simple ideas persist and are appealing? It isn't clear? That an art improves?
I think slop-pocalypse misses one thing. AI's time of thinking is very different than ours.
A "thick" argument that could take a lifetime in human terms could happen in minutes of AI.
Liking this concept - definitely seems to have more design freedom and ability to iterate than say Claude Design and at a lower price (well it doesn't blow through it's entire "token budget" after 3 prompts anyhow!)
The "copy agent prompt" feature has what looks like a thorough prompt for the implementation, that can call back to DiffUI to grab assets and make SVGs and other things a FE dev would bug a designer for - nice touch!
Re copy for agent, I originally had a build step that was build into diffui itself, but what I found was that:
1. People wanted their own agent with its own skills/extensions to do the work
2. People wanted to build in their own codebase / style
The copy for agent was a way to allow for that. I do have a MCP as well, but I find I actually really enjoy the process of doing visual exploration on the canvas first, then handing it off to an agent after.
Agree with both points, at least for folks who already have that infra (which I do). Made it much simpler than trying to "extract" things out of a tool that wanted to do everything.
To be fair to them, the architecture description said that each datasource had a unique agent, so the orchestrator AI didn't have direct data access, and that they specifically only allow access to data the user has permissions for.
Unclear if each datasource agent is ALSO AI based though, in which case it has just pushed the same concern down the line one hop.
Pretty much any Mac bought in the past 5 years can fulfil the requirements, which doesn't feel terribly unreasonable, and I bet the Intel case would be straightforward to cover too, and now you're catering for every Mac bought in the past 6 years.
Apple are dicks about making it easy to test on older macOS revisions, but I'm sure that'd actually be easier than you'd expect too. (I have a FOSS project that has macOS as a supported target. It targets OS X Mavericks, released in 2013. I don't have any easy way of testing anything older than macOS Ventura, released in 2022, and to be honest I don't really care to figure out how to do any better, but, last time somebody complained about OS X Mavericks incompatibility, and I fixed the problem, which was actually very easy to do: it did apparently work.) Put in a modicum of effort and I'm sure you can make this thing work for every Mac sold since like 2015, and there'd be a non-zero chance it'd work for some older ones too.
Thinking back to when BBSs were a thing, since that'd be on topic: perhaps Americans got a lucky break with the Apple II, in production from 1977-1993 (says Wikipedia) and seemingly a viable commercial platform for a measurable fraction of the period? For me, growing up in the UK in the late 20th century, the whole computer ecosystem seemed to get pretty much completely refreshed about every 10 years at the very most. Buy a BBC Micro in 1983: platform dead by 1990. Buy a ZX Spectrum in 1983: platform dead by 1991. Buy an Atari ST in 1985: platform dead by 1992. Buy an Amiga in 1986: platform dead by 1994. The PC was a bit of an exception, as the OS remained the same (for good or for ill...) for longer, but the onward march of hardware progress meant that you'd need new hardware every few years if you wanted to actually run recent software.
Anyway, my basic thinking is that if in March 2026 you are releasing some software that requires you to have a computer manufactured at some point in the 2020s, then this is hardly without historical precedent. It might even simply be the natural order of things.
Me? I set the displays to go to sleep after N minutes.
The great promise and the great disaster of LLMs is that for any topic on which we are "below average", the bland, average output seems to be a great improvement.
I think the "messy ideas" was a reference to the homepage copy "Turn your messy ideas into crystal clear specs.", not continuing the previous thought about the placeholder. I'd agree that "messy" might have more negative connotations than you intended.
Learned a few things I didn't know about exception handling, like Vectored Exception Handling. If it's possible to somehow have enough permissions to install a generic vectored exception handler that has enough complexity to emulate generic instructions, not sure why the shellcode couldn't just be included there instead.
Maybe someone else will have a follow on regarding some product that does some more complicated processing in a VEH that could be used to implement something that has the same shape as this.
TLDR: AI and generic pop culture artifacts have something in common - a lack of a certain type of depth that makes art great.