I note that even a hybrid kernel would require a compatibility layer to provide all the userspace stuff including the the display environment, access to the Unix file system, etc.

As far as compatibility layers go, years ago there was one called Merge which supported win32 on Unix. It was a commercial offering. I actually used a derivative of this back in the windows 95 days to run Windows 95 apps on Linux. Sadly they never made the transition to supporting modern windows NT apps.

Hint: You cannot run win32 apps on a unix kernel without a compatibility layer of some kind! (Which is what WINE *is*.) There's no way around this to get seamless win32 binary support. You either need to have WINE integrated into the OS and desktop environment, or some other implementation of win32 in the OS that translates win32 calls into kernel calls and unix libraries for I/O, screen display, etc. Or some kind of hybrid kernel.

All of this is true whether you're running windows apps on *nix or wanting to run Linux apps on FreeBSD, or Linux apps on Windows. What do you think WSL is?

Tight integration between WINE and the OS and desktop environment is certainly possible. If Wine were actually "finished" then I'm sure distros would be doing this. Meanwhile only MS could produce a true 100% win32 compatibility layer for Unix OS's.

Right. You mean with a compatibility layer like WINE. And that's literally what WINE is. An implementation of the win32 platform for the Linux (and now BSD) kernel.

Obviously. But the agents will suggest an appropriate language if asked. I suspect Rust is being recommended more and more.
And agents regularly do make arbitrary decisions if you don't explicitly tell them what to do. For example I know someone working on a large project and the agent chose python and some Tk-based UI library without being asked.

And like I said, it doesn't matter if I know Rust or not anymore. I can simply choose it (because it's popular, or because it is "safe") and the agent will go to work.

With more and more software development being turned over to AI coding agents, in many respects the language itself no longer matters. However I'm not surprised Rust is growing in popularity, driven by AI code generation. And it's a good match for that. While a lot of human developers find Rust difficult to learn and work with, AI agents have no such difficulties and the Rust compiler guarantees help mitigate some of the problems from mediocre, generated code.

I've seen quite a few projects introduced lately that all use AI code generation and Rust. And frankly that's probably the only way I'm going to use Rust myself.

Very few Android phones have home buttons these days. Usually they just have the power button and volume buttons. Home is either a swipe or a soft, on-screen button.

Not CodeWeave either.

Can't remember the last time I saw a reference to Earth Final Conflict on slashdot. Well done.

Just the other day I was thinking that at this point I'd probably welcome the Taelons with their disruptive technology (free energy, transportation, etc), even with their ambiguous and dark motives. Can't be much worse than what the people who've seized power in the three rival empires want to do to us all. Plus all the regional warlords (soon to be anyway) jockying for scraps of power. Eliminate the need for oil as an energy source would really take down their power and pride.

Thought-provoking fiction anyway. Which was what Roddenberry was always about.

At first I was concerned when I thought that CodeWeavers was in trouble. They are a plucky company that's the driving force behind much of the current Wine project development. They do a great job and we all benefit from their work.

Then I realized it was "CodeWeave" whoever they are. Some Cloud/AI company I guess. The only thing that's concerning is that the people who lose their shirts when companies like this one go under are often middle class people like you and me with their retirement money in managed funds. Rich people as a rule never do anything with their own money. They always use and lose other people's money. Especially in the US, the wealthiest and most powerful people always get someone else to pay for things they play with and even pay to work for them.

It is a relief that it's only one one bank that has a large number of security flaws,
and not all of the major financial institutions, all at the same time,
That would be awful.

I credit the steady hand of leadership for guiding the ship of commerce thru the Straits of Mythos.

Ahh, yes the UID checks out. Most slashdotters these days probably don't even know who Bill Gates was.

Yes there's a cost to continue doing business with companies that are increasingly greedy based in a country that is increasingly hostile and an existential threat. I have never on my life seen such brash and naked lust coming from a country that was previously a beacon, a country that built alliances that benefitted more than just themselves. Sad that this has all been destroyed. Hard times ahead for everyone. Sad also because it's so senseless.

Some people really enjoy getting beaten by the chain of command.
Takes the load off all that thinkin'.
Here, have more AI.

Nobody at Anthropic knows how it does what it does anymore, not really.
They can ask it and it will tell them things about itself,
but after just a few code turns, what was generated passes unit tests, it passes QA,
it's a programming miracle. Why bother looking at the code anymore?

And pretty quickly, no human knows how anything of it really works,
and if they want to make modifications, the code base is now out of the programmer's hands.
Someone (or itself) has to ask *it* to make the changes, forever,
because it holds the only architecture truth, and context of its revision history.

I used GPT to generate stuff, and yeah, it's quite remarkable.
But I also saw how seductive it was and soon, how after maybe three revisions,
I'd have to let it own the code and do all the work. I mean why not?
Any changes I make myself now brings the additional task of telling the AI what I did and why,
so it doesn't discard what I did at some (unknown) point in the future.

Other than the forever escalating cost of paying for a simulated employee
that you can no longer do without, it's amazing.

Do you ask to AI agent thoroughly document the architecture, APIs, protocols, and data flow as you go? Do you use planning mode in the coding agents?

Personally I have good success with coding agents when I spent significant time with the agent building out a precise specification document that describes in detail what I want to do, how it should do it, how the data is formatted, protocols to use, etc. And then I ask it to establish concrete step-by-step development plans. I'm only working on these projects in my spare time but I might spend 20 hours or more on this. Then after both I and the agent reviewer are satisfied do I start it generating code. Usually code generation is a tiny fraction of the planning time. The more precise the specification is, the faster and more accurate the agent can generate code to fulfill that specification. I review every code change the agent makes and I never let it commit to git without my authorization. Often I'll push back on a coding choice the agent selected and ask it for a detailed explanation, and often I'll tell it to do it a different way (it always apologizes for what it did tells me I'm brilliant for my suggested change when I do that but whatever).

This process may not generate code as fast as vibe coders crank it out, but it's more than twice as fast as I could do it, even with the long planning sessions.

There are some interesting youtube videos by "No Place Like Localhost" on this type of spec-diven agent coding.