this gordian knot of Anthropic and the USG is simple enough to cut in half. focus on the opensource agents training models outside of the state. right now, the winning move is between the cryptography that secures data which attackers target, and the environments which have no choice but to yield that data to encryption which doesn't leak. there still isn't a complete futarchy that drives either beyond some currently unmeasured risk. https://x.com/a16zcrypto/status/2021348447656476841?s=20 https://x.com/yinglun122/status/2026049333364507043