It goes without saying I'm magnitudes more """pro-AI""" than this website's median bloody-minded luddite, but it is fascinating that part of the appeal of having LLMs do web searches for you is that having a power hungry, computationally intensive mathematical miracle hosted however many hundred km away fetch the page for you, read the raw code, generate analysis, and then stream output to back to you is on the whole a faster way to retrieve information than fucking rendering JS in-browser.
And then webdevs have the gall to go and block Claude Code or whatever from downloading their websites because "ohhh 🥺🥺 so much downloading is massively costing us so much in bandwidth 🥺 we just can't afford it as if they can't not but send 150MB of CSS and React gibberish.
my 0.5 GB substack page #my0.5GBsubstackpage

















