i've written a harness for cybersecurity work that's like actually being used in real work by people who actually pay for it which feels kinda cool i'm working for a very very big company that makes security software, who, as with most of the tech sector, is very pro AI. thousands of lines a day are passing over us where the vast majority are at this point written by AI, i have sat sessions with major AI vendors where they gave us obviously faked demos. i get a lot of LLM exposure, like a lot. in very big production codebases. part of my specific responsibilities here are to try find places where LLM's can actually yield efficient and reliable positive-security outcomes, which is in far fewer places than they would have you believe i've never really liked AI, but when I got this responsibility it was important to me to do a good job at it. So i studied high-performance vibecoders and how they work and tried to integrate that wherever relevant and actually got some decent stuff done all the while, i would be frustrated when people in meetings would say shit like "can you ask Claude how we-", because, like... just think? over time though, i started finding AI a lot more useful, and starting to feel a little more positively towards it. every time I spoke to someone I'd hear people talk about how much better they were, how they were using LLMs to better themselves and their work they weren't talking just about productivity but also like actually being smarter, which was in many cases just observably not true. I was learning a lot though. as i started getting more used to LLMs I realized that there were more uses for them than I had previously guessed one day, i was just kinda bored, so i decided to do a lil binex challenge, nothing too interesting just something to pass the time, and as i got started, i opened a claude window and asked it to take a look that's when it kinda clicked that something had gone very wrong, and that reaching for an LLM had just become an ingrained behaviour i started looking back on how i was thinking about others, how they were obviously getting stupider by the day and how i was obviously doing something they weren't because i was smart as shit actually, and i realised, i was just as dumb as they were as someone who loves code and cybersecurity because of the potential for infinite learning, the realization that i had actually given up all my joy unknowingly and become stupider legit had me crying at my claude code so i decided i would get away from LLMs as much as possible. my job requires me to use them a lot so that couldn't really go anywhere, but twitter won't stop talking about 'em so i left here, i also put in a week PTO with the intention that that week, i would not touch an LLM even once, not even search engine auto-answers. where i would just _do things_ w my own stupid hands and learn as much as i could for the simple joy of having learned it without regard for applicability or even if anything i did was even good the whole time i kept thinking "fuck computer, everything is human", because I had a tweet from my moot @luxmendax stuck in my head the entire time i was trying not to lose my mind and decided that would be the theme for the week. SO - I read some books! - I downloaded Reaper and made a song! - I wrote some code which solves literally no problems - I played videogames - I did a patreon course thing from one of my fav creators - I did a bunch of CTFs I had a really really good time, and this time, i actually did learn a lot. The CTFs were fucking hard tho it was so easy to see the impact LLMs had on my brain. Challenges that used to take like 5 min were taking me 20+ in the beginning. It all came back pretty quick but HOLY SHIT that was rough to claw back. the reason i wanna write this down is bc i know i got homies in similar places atm, w frustration abt LLMs being used as first order optimal solutions, with LLMs taking the fun out of doing the things you love doing, but also, i know a lot of people are on here like "i've learned so much w AI". i'm sure that is the case for like... 1% of u... but the far likelier case is that u have alr lost more than u even remember having. there's no shame in that either, these machines are designed to make u stupid bc ur stupidity is dependence and dependence is profitable if u don't believe me and ur a person who does have daily LLM exposure measured in hours, try it, go a week without touching an LLM and still do all the things you're interested in now. fuck israel, fuck the police, fuck AI CEO's, fuck the computer everything is human thx