Anthropic’s Mythos is Back, OpenAI Releases GPT 5.6, Apple’s Price Increases
Original source
Guest
Ranjan Roy is a tech commentator and co-writer of the Substack Margins.
Summary
Alex Kantrowitz and Ranjan Roy discuss a week that underscored how frontier AI access is increasingly shaped by government and company-controlled release gates. Anthropic’s Mythos/Claude block was reportedly lifted for more than 100 U.S. institutions, while OpenAI’s GPT-5.6 arrived as a limited preview for trusted partners, reinforcing the idea that frontier models are becoming tightly managed strategic assets. They debate whether restricting access helps with safety and anti-distillation defenses or instead pushes customers toward cheaper models, routing, and open source alternatives. The conversation then shifts to enterprise AI economics: benchmark comparisons, token efficiency, half-price model claims, opaque billing, and audited overcharges all suggest customers are learning to spend less on premium models. In the Apple segment, both hosts argue that recent Mac and iPad price increases are too large to be explained by DRAM inflation alone and likely reflect deliberate margin expansion, with broader implications for device adoption, services revenue, and even succession strategy around John Ternus. The episode closes with a tribute to Om Malik and his influence on tech journalism.