By now, China should have already obtained critical AGI secrets. @leopoldasch once predicted that within 12 to 24 months, we would leak key AGI breakthroughs to the CCP. China's AGI labs won't be in Beijing or Shanghai—they'll be in San Francisco and London. AGI secrets are the most important national defense secrets of the U.S., but the way we treat them today is no different from how we treat any random SaaS software. You don't even need to launch some dramatic espionage operation to steal these secrets: you just need to attend any party in San Francisco or take a quick glance through an office window. I think China has a pretty clear path to entry: outbuild the U.S. and steal the algorithms. The main content comes from notes the author compiled during a board meeting at OpenAI, so the credibility is very high.