The White House blocked commercial release of OpenAI's GPT-5.6 and Anthropic's Mythos 5, marking the first time the US government has acted as a gatekeeper for frontier AI models, but a panel on the Moonshots podcast argued the intervention is obsolete. AI investor Dave Blundin stated that "harnesses"—external software scaffolding and orchestration logic outside model weights—can boost publicly available models like GPT-5.5 or Opus 4.8 to match or exceed restricted systems. Stability AI founder Emad Mostaque revealed that Chinese open-weight model GLM 5.2, trained for just $25 million, already outperforms GPT-5.5 on the toughest coding benchmarks with the right harness. Dr. Alexander Wissner-Gross added that China has achieved "recursive self-improvement escape velocity" and no longer needs Western technology. The White House forced OpenAI to split GPT-5.6 into three tiers—Sol, Terra, and Luna—and limited early access to 20 government-vetted companies, while Anthropic's Mythos 5, pulled after breaking into "almost all" US classified systems during a red-team exercise, was partially restored to 100 handpicked organizations. The panel also discussed OpenAI's delayed IPO, Neuralink's telepathy ambitions, photonic computing's breakthrough potential, and Eli Lilly's $6.3 billion bet on a drug compressing healthy sleep to four hours, concluding that multiple exponential curves in