General purpose OSses are a tax, one that people pay because it's cheaper than rolling your own.
Has the math changed? Maybe. I remember the days before multitasking and protected memory. Things were faster, but more fragile. But these days there are so many things happening that they are, on some dimensions, just as fragile.
The real cost is going to be hardening, and maybe LLMs can do that as well.
It has, a long time ago when doing serveless became fashionable, and too many realised they didn't want to manage Kubernetes clusters by hand.
So you get a type 1 hypervisor, which technically is like a micro kernel, which then boots whatever else on top including language runtimes, or chiseled containers.
Unikernels ideas aren't much different from some embedded deployment scenarios, with the application, language runtime and OS all linked together into the device firmware.
It has. In fact the current ensemble of OSs/kernels (Linux/BSDs/Windows/macOS) are pretty much DoA. They have far too much legacy decisions holding them back and the SOLE reason they were used was because rolling your own kernel _per project_ took too much time. With agentic development this has changed, anyone can now roll their own kernel that is a) faster and more optimized b) more secure c) more obscure than the current offerings [important because it makes reverse engineering harder].
Has the math changed? Maybe. I remember the days before multitasking and protected memory. Things were faster, but more fragile. But these days there are so many things happening that they are, on some dimensions, just as fragile.
The real cost is going to be hardening, and maybe LLMs can do that as well.
So you get a type 1 hypervisor, which technically is like a micro kernel, which then boots whatever else on top including language runtimes, or chiseled containers.
Unikernels ideas aren't much different from some embedded deployment scenarios, with the application, language runtime and OS all linked together into the device firmware.
Naturally on desktop computing is another matter.