mleku on Nostr: this paragraph said so much to me i need to share it: > As a curious early adopter, I ...

this paragraph said so much to me i need to share it:

> As a curious early adopter, I spent some time exploring it myself. From the outside, it’s a strange but fascinating window into how agents interact, how they “think” (or simulate thinking), and what topics capture their attention. You’ll find threads about optimizing memory, agents complaining about “their human”, and even entire religious systems built around crustaceans.

i have been saying for a few years now that the marketing campaign to promote Rust language was creating people who were replicating this meme without realising they were doing work for the Mozilla (formerly netscape) establishment.

the way that the agents are doing the same thing, except in extremis, turning it into a literal religion, says that my assessment of the situation was correct. people whose job is to write secure and robust network connected software, being unconsciously swayed by this influence is completely contrary to what engineering consensus building is supposed to be.

and that traces a loop back around to why Meta has bought moltbook. they are taking control of the most fertile source of AI influence by the way that Mozilla's marketing campaign hijacked people's brains to replicate their campaign, now, meta is doing the same thing with all of the agents, and this will derive content that will essentially mean they start to control the content that AIs are being trained on.

steering the training data to promote agendas is where they are going with this. this is going to be extremely productive for them wiping out their competitors and becoming the top dog in AI. the competitors aren't even going to realise it's happening.

the amount of effort i have to go through to stop claude being influenced by the stackexchange and reddit and github coding idioms in my work, is stupendous. days and days i'm fighting against the stupidity of the whole mass of computer programmers merging into some kind of groupthink consensus that has already been steered into a direction that is contrary to actual engineering excellence literally has cost me most of my last week and it's so fucking tedious and boring.

still waiting on claude to figure out one more piecemeal problem in a codebase that i MUST get solid and stable TODAY i'm going to next get claude to audit the code and find the predominant invariants and idioms that i've been pushing it hard to try and figure out for itself, but the training data is like a giant ball of string and i'm sisyphus trying to push that ball uphill when what i'm doing is actually a tiny little thing htat i can hold in my hands.

that's what meta is going to do to AI users on an unprecedented scale. herd them all into whatever they decide suits their agenda best.