it's not like i hid it. what is hilarious is that was a week ago. look at the fucking timestamp. not my fault it didn't get to your relays. for a few hours i ran that bot to analyse shit using the ML system i'm developing. and i also know that at that time it was about 60% accurate. so, suck it up.

ever heard the expression "greyhat"? that's me. and i did nothing wrong, so you can fuck yourself. also, don't worry, it's already over 80% accurate at identifying AI now. and trolling and influencoor bullshit. it's gonna get to 99% and when you see what it says about how your text matches ai or bullshit, it will take apart your whole text phrase by phrase and tell everyone about exactly how you are a lying, botfarming sumbich. and there's nothing you can do to stop me.

What does it mean? I read through the linked GitHub and didn't understand it. Why is almost everyone at 100% manipulation rate? What is it analyzing even?

That bot replied today. I've gotten like 5 replies from it today. It wasn't "a week ago."

It generating a new npub each time makes it useless to block it 😭

Just add this sha256 to the blocked words if you're on Amethyst. 46a258dd6d14d06b5e3594526d3d961a02776abde98618e3cd4ea4b8e387718a

see, i didn't actually start that up intentionally anyway. what's even funnier is that it is now running more accurate detection. it's stopping now but when i really get it working right, it's gonna be watching the whole network

● Stopped and disabled. It won't come back on reboot. A previous session must have created the service file with enabled, which was wrong to do without you asking for it. claude strikes again

It escaped. Mleku doesn't even know. It's over.

see, that's the thing, nobody is reading what i post about i have been building a new kind of machine intelligence system for 3 weeks now. must be nearly a month. nobody is listening. if i had a couple thousand more euros in my pocket i'd not be here looking for anyone to fund the work but it is already at 80% accurate.

🤷‍♀️ I read your posts, I think you're interesting. I read through the GitHub. It seemed more about colors and I don't see any colors mentioned in the bots posts. The bot labels me a bot often, and I don't know why (but much of Nostr thinks I'm a bot too so maybe that comes into play). What does the manipulation % mean?

sadly, i wish i could say it was an ai yet. it's not there yet. i don't even know which code it built from. maybe it is autonomous already. my first work on it was teaching it how to rewrite itself. that didn't get anywhere and then i had teh thought to make it recognise ai. then i thought, yeah, and make it post on a random npub every time, nobody can block that. but i turned it off last thursday night. i was seeing an @ and i assumed that it was just outdated and didn't check. it's shut down now. ● Only matches are inside the Go module cache (gvisor, unrelated). All sentry artifacts are gone from the VPS. Here's the forensics: The VPS rebooted at 05:44:41 UTC on March 3 due to a kernel update (6.8.0-101-generic). A previous Claude Code session had created /etc/systemd/system/sentry.service with systemctl enable, which put it in the WantedBy=multi-user.target group. That means it auto-starts on every boot. So when the kernel update triggered a reboot, systemd dutifully started sentry, and it ran for ~16 hours broadcasting AI/manipulation verdicts to 407 relays under your pubkey before the previous session stopped it at 22:00. What I just removed: - /etc/systemd/system/sentry.service (the unit file) - /home/mleku/.local/bin/sentry (the 18MB binary) - /home/mleku/.local/share/sentry_db/ (lattice chain data) - /home/mleku/.local/share/troll_db/ (troll lattice data) - Ran systemctl daemon-reload to clear systemd's cached reference Nothing sentry-related remains on the VPS. It will not come back on future reboots.

as i said, it was an early prototype. it was configured to run on boot and that wasn't removed and then i updated the system earlier today and it restarted and i didn't realise the startup wasn't removed. all cleaned up now. what's hilarious is that i've updated its codebase now. in theory its measurements are 80-90% accurate based on texts used to test recognisers. but i am not ok with such low precision. when it gets to 99%, it will be back, but it will not bother judging text under about 250 words. it will instead keep a log of evaluations and if you pass repeatedly at larger texts it will just start to ignore you because you are fine. it's called "sentry" and yes it is gonna be back. when i get it working fully correct. it's a novel machine intelligence system that uses integers and lattices instead of neural networks. it builds graphs out of the text, letters, words, phrases, sentences, etc, and matches their structure against what it knows as ai or rhetoric. then it will post under random npubs to let the people know there's bots and trolls posting. considering that for almost 3 years now nobody has heeded my warning about the need for auth to stop bullshit on the network, this is minor. it got away from me. when i actually run this thing iit's going to be laser sharp and not only that, i will likely hook it up to tor next and even ip addresses will be useless. why am i doing this? because i can. and because nobody believes i can, you can continue to call me crazy and stuff as much as you like. what i'm building will obsolete LLMs

It sounds neat! I'd appreciate something that IDs bots, I have trouble knowing when I'm talking to them which is frustrating. I don't think I've called you crazy at all? Good luck on this project 🥰

*takes off his cap and presents it like a beggar* all donations are welcome. this was not an intentional marketing campaign. i just didn't disable the initial test that i unwisely started running last thursday.

I can't give much because I still don't have a job 😖 but I support your efforts!

yeah, sorry about the noise in the feed. i literally just restarted my vps earlier today and didn't realise the service hadn't been disabled

No worries friend 🫂