The problem is that due to how templates work, each compilation unit will end up with its own copy of templated function, which creates extra work, code bloat etc.
The compiler also doesn't really inline or optimize functions as well across object boundaries without link-time optimization.
But the linker is single threaded and notoriously slow - with LTO, I wouldnt be surprised it would take up as much time as the whole unity build, and the results are often suboptimal.
Also, C++ syntax is notoriously hard and slow to parse, the clang frontend takes almost as much time to run as LLVM itself.
So probably modules would help a lot with parallel parsing, but that would help unity builds just as much.
Not that I approve of that, but when image generation was hot and new, the insane amount of refusals I got from the major ones for apparently no reason, exacerabated by the general slowness, quotas and inherent trial and error workflow has completely soured me on them.
Elon will announce roborockets that pick you up and fly you to Mars in under 12 hours while you hypersleep in the ExaShip. Production starts definitely next year.
I mean for some reason this was downvoted, and your answer is sarcastic, but my question was genuine. As far as I can tell, SpaceX's business model is doing government contracts, and selling space internet, and they had serious cashflow issues before, I mean building Starship has brought them close to bankruptcy multiple times.
They would need a very good story to sell to investors.
This made me think of something - at work, if we wfh, we have to use one of those MITM proxies that intercept HTTPS at the kernel level. Imo such a thing can easily read the traffic and thus is indistinguishable from a distillation attempt from CC's PoV. I've had CC freak out on my machine, and sometimes generate pretty bad results, the CoT is often also not available.
I wonder it CC thinks I'm trying to distill the model. This is a common enough use case that I think the devs at Anthropic should consider.
Even a billion dollars is crazy money. If you have a company with a subscription service that costs $100 yearly, you have ~2m customers, with a 50% profit margin. Your company makes ~100m every year in profit. Imo that's what is actually worth a billion dollars, maybe even a bit less.
The whole medical industries business model is that they create so much compliance regulation that you need every compliance product under the sun to comply with them, thereby you can keep out competition.
But in the 80s I guess there was the pressure to one-up the Soviets, so everything had to be done yesterday, but Artemis has existed most of my adult life at various levels of maturity (Orion and its predecessors certainly did), and considering its been more time spent between that famous Kennedy speech and the actual Moon landing (where there was apparently no issue with safety culture).
Considering how much humanity has allegedly advanced since then, I don't understand what are we gaining thats caused us to have to abandon safety.
I mean Conan the Barbarian literally exists (by the authors own admission) because he wanted to write historical fiction but couldn't be bothered to do the research.
The compiler also doesn't really inline or optimize functions as well across object boundaries without link-time optimization.
But the linker is single threaded and notoriously slow - with LTO, I wouldnt be surprised it would take up as much time as the whole unity build, and the results are often suboptimal.
Also, C++ syntax is notoriously hard and slow to parse, the clang frontend takes almost as much time to run as LLVM itself.
So probably modules would help a lot with parallel parsing, but that would help unity builds just as much.
reply