Anthropic Launches Claude Opus 5.5, 40 % Cheaper with New Cyber Safeguards

Capability leaps arriving between routine versions recall the 1949 Soviet atomic test that blindsided U.S. planners: progress came sooner—and quieter—than models predicted. Opus 5.5’s better-for-less economics compress the cost curve that Moore’s Law stretched over five decades into a bi-monthly cadence, hinting at an AI variant of Wright’s Law where every doubling of output slashes price. The new fallback scheme echoes early arms-control norms like the 1975 BWC—voluntary, porous, yet a starting signal that raw capability now demands shared restraint. Commercial pressure (imminent trillion-dollar IPO, price war with OpenAI and Chinese open-weights) is already overriding the CEO’s recent call to “pace the frontier,” suggesting that economic incentives, not pledges, will dictate tempo. If the 1980s PC boom democratized software creation, 2020s rapid-fire model upgrades could democratize exploit creation; a century hence historians may mark this release as evidence that frontier AI entered the mass-market phase before governance caught up.