>AI agents start using wget and curl to download webpages and read themwhere were you when you realized everything is circling back to the GNU Philosophy
>>108891373as opposed to what? opening a tcp socket to the web server?
>>108891455as opposed to using MCP slop or other obtuse APIs to retrieve the datait's literally a RETVRN to the command line
Tools and libraries are on a borrowed time, eventually AI will just directly generate programs for specific tasks.
>>108891663agy can just directly use chromehttps://antigravity.google/docs/browser
>>108891681Praise Hashem
My agent has a tool especific for reading websites, it wgets on the background, extracts text elements and returns them. The smallest interactive subagents get a simplified version where only interactable components and nearby text is given to them.>>108891681It already happens, I have to regularly clean the folder for 1-time-use programs, if it sees an error it also refuses to edit the file and writes it all over again on another file.
soon we'll be piping agents into each other
>>108891681>Tools and libraries are on a borrowed timellms are on borrowed timeprofits when faggot
>>108891373Don't they have privleged API access? Obviously being at the top of the pecking order of big tech investment. When I see ChatGPT doing "Searching web..." I assume it has hard coded privileged access.
>>108891760Where we're going we won't need profits, Jimbo.
>>108891799The fuck are you saying. They just use goggle's paid API shit. You can do it too. It costs money though. So usually you use the free version on the web, the one that is filled with aids and does not really find your shit.
>>108891809I mean when they scrape websites, they may have a deal with them to allow to scrape them with no blocks. Since the web is teeming with scrape bots a lot of websites will have mitigations against it (proof of work is a typical example), so I imagine ChatGPT makes deals directly with sites to let them scrape.
>>108891814Oh, the probably do those, but only with select websites. The only "big" thing related to paid scraping (web scraping loicense) was that cloudflare API thing. And it's pretty bad.
>>108891799they have some service apparently but half the time i just have them write a python script to scrape it with playwright so i dont have to rely on it for anything
>>108891455The AI agents are written in javascript. They could just download the page with javascript.
>>108891977codex is written in Rustits only anthropicslop written in javashit
