Intelligence Buying Intelligence (stevekrouse.com)

Recent discussions in the AI/ML community highlight a paradoxical trend where users continue to opt for slower and more expensive closed models like Opus 4.6 over faster and cheaper open-source alternatives like Sonnet 4.6, which is nearly twice as efficient. The author, Nnamdi Iregbulem, argues that trust issues, stemming from frequent mistakes in less expensive models, drive this preference, as users are wary of the catastrophic consequences from relying on less intelligent systems. He emphasizes the non-linear nature of intelligence—a slight edge in capability can yield significantly better outcomes, similar to how top investors like Warren Buffett leverage unique insights to achieve success. Iregbulem explores the implications of intelligence as a “buying” commodity, noting its increasing costs due to demand for higher-performing models and efficient computing. He introduces the concept of the "Lawyer Flippening," where advanced AIs become strategic partners, routing tasks to ensure optimal use of intelligence. This reshaping of roles means that in a future dominated by smarter AI, users must learn to allocate tasks effectively, leveraging the best available intelligence while grappling with the challenge of managing potentially high costs. As AI models continue to evolve, it raises critical questions about how we balance the investment in intelligence against its practical applications and economic returns.