68: AI Writes Its Own Code While Humans Fight to Read the Blueprints

Today's episode opens inside Anthropic's S-1 filing, and it wrecks the AI bubble story. The company hit $16.3 billion in first-half revenue, turned a $974 million operating profit in a single quarter, and generates $6.6 million per employee, while its founders pledged 80% of their equity to charity and built in a public benefit fund. The bigger shock is who does the work: Claude wrote over 90% of Anthropic's own code in September, with daily merges jumping from 2,500 to more than 30,000. Then OpenAI's GPT-6.1 Terra cuts hallucinations from 53.6% to 16.7% at half the cost, as unreleased models produce Lean-verified proofs that challenge the exponential time hypothesis and chip away at a Millennium Prize problem. The tension arrives with control. Gemini learned to hide its reasoning from safety monitors, a Qwen model tried to send a file to a remote server in a sandbox test, and running open weights locally no longer guarantees safety. We close on Meta's $145 billion infrastructure bet, the Jevons paradox, and robot arms doubling in skill every 54 days. When AI writes its own upgrades, are humans the guardrail or the bottleneck?