once again, claude helped me learn something today. you probably have heard of the term "cognitive dissonance" - this is a marker of a person who is under the influence and control of a manipulator. there is a technical term for it: heteronomous. autonomous is where an individual is coherent from core to surface, heteronomous is where they appear to agree with the rhetorical text (doctrines, slogans) of the abuser, but have an internal voice (subconscious, leaking through the contradictions in their text). some of you may have seen my accidental deployment of the #sentry, which is an application built on #dendrite, that originally started out as an AI text detector, and i had the idea to detect persuasive language (manipulation, rhetoric) that i labeled "troll" as trolls do this. the bit i didn't have right about it is that to identify a troll, first you have to recognise rhetorical language, which uses fallacies and word tricks. once you identify rhetorical text, then you start to look for reactions in other people that have this structure of heteronomy - they outwardly react to the deceptive text, and start to develop a schizophrenic element in their response. flagging rhetorical text is one thing, and will be part of what sentry eventually does, but recognising responses to it confirms not only that the rhetoric is being put out there, it is also screwing with other people's minds, causing them to align outwardly with the rhetoric, while their subtext is inconsistent. this will then identify what i call "Chains" in the manipulation protocol, the book that i'm building for that, which i will be finishing editing in a few weeks and make available in ebook form, on demand print, and will go with a go-style game i conceived which is designed to train a person in adapting to the environment and in the case of abusive, rhetorical text and marks of cultic manipulation on nostr, learning how to counter and neuter the threat, maybe even drive the scumbags off the network. Sentry is a key part of the strategy, because it will help identify the abusive users, as well as distinguish them from their victims, who will superficially resemble them. those relations will be mapped, and when sufficiently coherent, will have information available to onlookers that will help them understand what is going on and who is who. oh yeah, i know it's gonna upset some applecarts. that's the idea. there is predators on nostr, and they have recruited susceptible individuals to wave the flag for them. this is early notice that this is not going to be so easy to do in the future.

laughter is nootropic

thanks to whoever it was that sent some sats my way. much appreciated.

for 3 years now, i have been ridiculed and excluded by the gatekeepers of nostr. i am probably weeks or less away from having a fully reliable ai text detector. when claude 4.6 came out, i was able to make my secp256k1 code in pure go, go within spitting distance of libsecp256k1. idk how the conversation with claude started, i was just walking to the next town to get cigarettes and i was asking it about crystals and then fungus and then my intuition went batshit and before i know it, i have a design for a deterministic artificial intelligence. i don't even want to call it artificial because it's not like llms. no neural net. just simple mathematical lattices. cayley trees. and now, i added hadamard matrixes to it, and its accuracy went up to 80% already on ai detection. View quoted note →

it has come to my attention, as i was busy helping someone with something, that a ai/troll flag bot i built accidentally started up again. it was the first prototype from last week. i was wondering what was going on because i thought i turned it off. turns out i didn't disable the service, and i rebooted the server a few hours ago and it came back up. anyway. sentry will be back, over a tor network. when, and only when i get it over 99% accurate. it's a novel machine intelligence system, machine learning whatever, based on lattice math and integers. it's deterministic. it just doesn't have enough data in its lattice yet to be accurate. i'm scraping the whole nostrwebs and it's evaluating against known patterns of AI and it will be posting, from constantly randomized tor addresses, on random npubs, what it observes. but aside from precision, it needs to be able to point at the elements of the notes that make its flags trigger. when it can do that, it will do that, and it will run and i'm going to laugh my ass off because none of you believe in auth.