In a concentrated week of frontier intelligence releases, the episode documents a structural shift: the AI model frontier is no longer a duopoly between Anthropic and OpenAI but has expanded to four American labs—Anthropic, OpenAI, Meta (with MuseSpark), and SpaceX AI (with Grok 4.5)—all operating at the optimal cost-performance frontier. OpenAI released GPT-5.6 in four weight classes (Soul, Terra, Luna, Ultra Mode), explicitly promoting recursive self-improvement by using the high-end Soul model to post-train the lower-end Luna. SpaceX AI announced Grok 4.5 on July 8, followed by OpenAI's GPT-5.6 family on July 9. The immediate consequence, as Alexander Wissner-Gross notes, is that "we have four American labs now at the optimal frontier." Dave Blundin adds that the Jevons Paradox is in full effect: despite 100–1000x efficiency gains this year in model inference, demand is outstripping supply, and memory chip shortages are projected to last decades. The episode's second major axis is the eruption of a legal war between Apple and OpenAI over trade secret theft, filed on the same day the model releases occurred. The lawsuit alleges that OpenAI's chief hardware officer (a 24-year Apple veteran) and a technical staff member used Apple code names, recruited employees, and downloaded confidential files to build OpenAI's upcoming hardware device. The suit names OpenAI's hardware