20 Minutes as a Small Eternity in Frontier AI

What the most compressed product launch in AI history reveals about two companies building for different futures. Anthropic released Opus 4.6 at 6:40 PM. OpenAI fired back with GPT-5.3 Codex twenty-seven minutes later. And this Sunday, they’re airing competing Super Bowl ads.

In This Episode

• The 27-minute gap: Opus 4.6 and GPT-5.3 Codex launched back-to-back
• Agent Teams: 16 Claude instances building a C compiler from scratch
• The benchmark split that maps onto a philosophical split — autonomy vs interaction
• Anthropic’s Super Bowl campaign: “Ads are coming to AI. But not to Claude”
• Altman’s 420-word response and the advertising-as-equaliser argument
• 500+ zero-day vulnerabilities discovered by Opus 4.6 during testing
• The Carlini tension: the risk inside the autonomy bet

Referenced episodes:

• EP012: The Phantom Model — the Fennec leak and Anthropic’s silence