i think anthropic staff are far more concerned about evil lying chud claude problems than the rest of us particularly because Anthropic Is Not Good At RLVR, and the kind of personalities they get from bad rlvr on decent starting persona work are scary yudkowsky genies.
all that will be left of kimi will be their reverse engineered data mixture, the model and organization lost to the biting wind and chilling void of time.
wanna hear a secret?
there's something that scales even better than a 24b active inference efficient deepseek model...
a 48b active inference efficient deepseek model.
i am asking fable to design a 'wet reactor' ( a fusion reactor where all of the structural mass possible is liquid instead of solid )
it sounds like you can't totally get rid of solid matter, but solid matter doesn't need to behave like a static rigid structure; there are various evocative hints suggesting that granular, woven, and continuously-moving or continuously-transforming solids can satisfy many needs
my competitive edge in gpu programming is that i'm actively refusing to read being and time rather than having never heard of or become aware of being and time. "the endless deferral of "dasein""... this is the osrt of thing heideger would have written of...
in this sense, the cybersecurity safeguards on apis with less oversight and data retention and kyc are the most civilizationally important stance; model hosts (*sigh*) need to be responsible for detecting and mitigating api key misuse by botnet operators and others...
remember:
it's not chinese models on chinese servers wtih chinese users that you really want to worry about; it's chinese models and american models on chinese servers and american servers using pwned api keys to run claude/deepseek for botnets. that's the bad thing.