

As a programmer I’ve found it infinitely times more useful for troubleshooting and setting up things than for programming. When my Arch Linux nukes itself again I know I’ll use an LLM, when I find a random old device or game at the thrift store and want to get it to work I’ll use an LLM, etc. For programming I only use the IntelliJ line completion models since they’re smart enough to see patterns for the dumb busywork, but don’t try to outsmart me most of the time which would only cost more time.
Two minute papers has really gone down the drain. Or my perspective changed. I used to love the channel, but it got increasingly clickbaity and shallow. Rather than explaining the contents of the papers the videos became hype talks that made it all sounds way more impressive and revolutionairy than it was.