to all the managers/c-suite delighted with the idea of replacing all the workers by LLMs, remember: once you do, don't get surprised when your customers decide to cut the middle man and replace you and your company by some prompts. To those that believe they're so brilliant: don't underestimate an (inevitable) army of fully LLM created companies trying different combinations until they mop the floor with your Steve Jobs 2.0 complex by (evolutional) brute force

no matter how easy it is to type a password, with enough use you'll manage to screw it up so badly that you'll have to type it very slowly

patiently waiting for debian zurg to be announced

@stavros after forgetting my solokey again at home was pondering making a framework module for it and sure enough someone had the same idea and executed it better than I could. Thanks!

My top thing I'd like LLMs to solve is to generate code to allow using properly multiple versions of libraries so we stop with the trend of shipping tons of copies of different versions of libraries. If a library project doesn't care about API/ABI stability, do you believe they maintain and backport security fixes for old versions? That, of course, if we can ever trust LLMs to write code.

@bookwar @kernellogger Writing a gitforge from scratch in a way kernel people would use is infinitely more work than evolving the current process slowly. Or do you think Linus would put all the eggs in someone else's basket ever again after bitkeeper? As for CI, current model of "third party" CIs checking stuff without getting in the way works better than hoping some team will keep maintaining those tests and dealing with transient failures, hardware and toolchain failures.