CMake

I don't know how I feel about that.

At least the example that you're doing here is CMake.

And if you include or remove files, then it should change the behavior of the configure phase.

At least

you can still trick it.

So one example.

The reason why I brought up CMake though is that best practice in CMake is generally when you're defining a target, so executable or a library, generally you're supposed to hard code in the library paths.

And so then it will automatically do a CMake reconfigure if you add or remove a file, right? Whereas if you do something like globbing, then it doesn't actually become aware of when like additions or removals occur.

You write a build.zig file and you get a cross-compiler that can target any platform from any platform.

You don't need a build system generator like CMake.

You just run zig build and it works.

That's a huge deal for maintainers who are tired of wrestling with build configurations.

So, This would be like a turn off for you, I guess.

Because you need to have a CMake-based project, basically.

You can't just have your

own build script that CLion will call?

But yeah, a lot of stuff is possible.

Until earlier this week, for all of this to work, you had to first pre-initialize your build folder with some CMake magic so that CMake would have some metadata available to use the so-called CMake file API.

I think that's what it's called.

And so you have to do this first and put some, yeah, whatever, do some magic there.

Now, given that we use a slightly newer version of CMake, this is not needed anymore.

So you don't have to do west spdx dash dash init anymore.

You can start with a vanilla.

And that would be good as a feature for me to learn, and it would also be good as a way for me to understand what's happening with all this Gen AI software engineering stuff, because Ultimately, I felt as a leader, I've got a large engineering function underneath me.

If I can't live that experience of how our careers are changing, how that role is changing, then how can I sort of lead for it? I started out super slowly, like trying to remember how Make works and CMake and all of this kind of stuff.

It was really, really painful.

And if you look at the commit history on GitHub, you can absolutely see that I was in pain last summer trying to work this stuff through.

Frontier models are impressive but expensive and constrained, while smaller, cheaper, and local alternatives are quietly eating tasks people didn't think they could.

Add in a mystery model on OpenRouter and someone implementing GPT-2 in CMake, and it's a fairly representative slice of where things stand.

Anthropic's premium model isn't converting to premium revenue.

The Financial Times reports that Anthropic's flagship Fable model, widely seen as technically ahead, is struggling to attract paying users while cheaper competitors clean up the mass market.

The audio and brain run on an M4 MacBook, while the game runs on Windows.

Commenters mostly agreed this is what NPCs should look like from now on, and speculated about dedicated AI silicon in future consoles.

And in the why not department, someone implemented GPT-2 as pure CMake, using Q16.

16 fixed-point arithmetic and relegating Python to build tool duty.

A quick mention.

Most people probably don't realize or necessarily care directly, but we bumped the minimum required version of CMake to sort of align with what's widely available in common distros out there.

So it went from 3.20 to 3.28. It's still super conservative.

I think 328 is still from three years ago almost.

But this actually brings us a couple improvements from the CMake project.

There's one that I was actually personally really interested in that will help make the SPDX SBOM tooling more straightforward.

I

Yes, exactly.

Everyone hates tar balls.

No one wants to run compilers from the command line and running configure and installing CMake.

Ain't no one wants to do any of that stuff.

I can say all of that and still say I hate Homebrew.