AI coding has hit an 'intelligence saturation' point, Cognition cofounder says (www.businessinsider.com)

Russell Kaplan, cofounder of AI coding startup Cognition, announced a significant shift in the AI landscape, stating that many software engineering tasks have reached an "intelligence saturation" point. In a recent post, Kaplan highlighted that as AI models become capable of performing tasks effectively, factors such as cost and speed are becoming more critical. This shift is underscored by a report from The Information showing that the advantage in model intelligence has decreased significantly; a 10-point advantage now yields only 20% more traffic for developers, down from nearly 69% in late 2025. This change could have substantial implications for leading AI companies like OpenAI, Anthropic, Meta, and Google, as they compete to attract software developers and enterprise contracts. With the increasing capability of AI agents to handle routine tasks, a marginal increase in intelligence may become less appealing compared to more cost-effective and faster alternatives. The evolving focus on procurement efficiencies could reshape the competitive dynamics in the AI/ML community, making performance and pricing pivotal in choosing coding tools for developers and organizations alike.