update 4 years later: i stopped supplementing zinc but ive increased my dietary intake significantly, mainly by improving the overall quality of meat in my diet.

RT @qorprate: zinc effect notes:.- inhibits estrogen production by fat cells (net effect: increased T).- seems to affect sleep-wake, making….

RT @qorprate: zinc tips:.- take it with protein.- don't take it with magnesium (or calcium).- i take 10mg orotate. some pills have 25-50mg.….

looks like the kinda thing that picked fruit from the tree of knowledge and offered it to housewife.

oh no claude code . what have you done??

otoh once they train one of those NRM sudoku solver logic understander things on code and plug it into a big chat model as an actual reasoning module, then we can talk vibecoding.

the amount of software engineering expertise required for me to "vibecode" the app im working on has me convinced that vibecoding anything beyond static sites is a hype bubble. i have, however, learned a lot of new things from claude and gemini in the process.

I kind of like how Claude Pro usage limits are basically like "ok dawg you've coded enough for this several hour window take a damn break".

RT @strnglft: The wait is over. After all these years, @qorprate and I have finally completed the End of Evangelion.

anyway massive hubris for them to call it AlphaGo Moment.

I built something like this once but for personality quizzes. It cost a lot to run and made slightly better quizzes according to my benchmark that were also completely useless to human test-takers, because the questions were difficult to give accurate answers for.

lemme tell you: if you could design an LLM with STREET smarts and then gave it a vending machine to run, it'd make BANK (and probably end up fueling the entirety of Anthropic's Adderall addiction).

the material of the LLMs knowing is in "pure" relational abstraction over language, which is identical to one form of knowing humans use ("book smarts"). humans can also know in other ways, though. any questions?.

anyway to weigh in on the object level: LLMs can know things like any other machine, because the knowledge is materially inscribed within it. just like how a washing machine knows how to wash clothes, so an LLM knows how to predict tokens.

tbf this year is probably the first time ever that many engineers considered the grounds of the possibility of knowledge in any detail, and I don't blame them.

worst part about AI season is how i keep getting sniped by career engineers making incoherent metaphysical claims about the nature of truth and knowledge.

tbf the original post is right that incels are choosing to be incels because it justifies externalizing their rage, but it also obviously sucks to have such an amount of rage, and doubling down on intentional aura-maxxing doesn't address the rage at all -- much the opposite.

the problem with this kind of 10 step program to de-incel you is that it just increases resentment of the "demands" being made. better to just give up on fantasies of sex for a while and learn how to feel valued + love others as a person among other people.

if i were an incel, I would simply take a vow of celibacy, transforming me into a volcel, then make a bunch of friends, let the joy of compassion into my soul to erode my rage, then do whatever I like.

Thanks for writing this up. I find the current state of intellectual property law despicable. Wishing the best of luck to Anthropic as they weather this storm.