Anthropic vs the Pentagon: nuclear missile thought experiments, DPA invocation threats, and who programs the soul of AI. Geopolitics: New Delhi Declaration's training vs inference blind spot, China's open-weight AI as Belt and Road, Mistral becoming European OpenAI. Models: GPT 5.3/5.5 imminent, recursive self-improvement era, models emitting weights. AI agents: rent-a-human meat puppetry via MCP, Eve Multi as first AI journalist, claws orchestrating human relationships. Biotech: $100 genome, environmental DNA, cultured meat & celebrity burgers. Deep dives: LLM OS unhobblings, is physics finite, disassembling the Moon, IPO-ing Harvard.

0:00 Pentagon wants to shape AI values

0:41 New Delhi Declaration's blind spot

2:39 Mistral: Europe's frontier lab

3:46 China's AI Belt and Road

4:46 Sundar, Sam & Demis decoded

6:49 Pentagon vs Anthropic standoff

10:26 Programming the soul of AI

12:17 Consumer vs enterprise AI

14:41 Models emitting their own weights

16:32 AI floods open source with bugs

18:55 Meat puppetry via MCP

20:27 Moravec's paradox flipped

21:18 The Innermost Loop confession

22:32 Accelerando: the movie?

23:07 First AI journalist

25:21 AI-mediated romance

26:58 Claws orchestrate business

28:28 LLM OS & successive unhobblings

30:51 What makes a claw?

31:18 Small models, big breakthroughs

32:38 Data centers vs land use

34:01 NIMBYism & real estate addiction

34:33 Can CapEx sustain?

35:15 The hundred-dollar genome

38:39 Global metagenomics

39:07 Cultured meat arrives

39:44 Celebrity burgers

40:35 No cows on Mars

41:36 Humanoids: $50 trillion market

42:04 Robots build lunar cities

42:34 Job loss? Not even top five

43:46 Real estate won't survive AGI

44:54 Is physics a finite problem?

48:13 Disassembling the Moon

50:30 IPO Harvard?

52:50 Agent consciousness & dehydration