Vertical AI's Bitter Pill: The Rise of Horizontal AI

2025-01-21

This post explores the competition between vertical AI applications (AI optimized for specific domains) and horizontal AI applications (more general-purpose, scalable AI). Using personal experience and Hamilton Helmer's Seven Powers framework, the author argues that as model performance improves, vertical AI applications struggle to maintain a competitive edge. Except for a rare few possessing exclusive and essential resources, most vertical AI applications will eventually be overtaken by superior horizontal AI. Horizontal AI, akin to a remote coworker, is easily integrated, cheaper, and continuously improves performance through model advancements. The author uses their AcademicGPT project as a case study, showing how a vertical AI application was surpassed by more general horizontal AI models.

Read more

The Bitter Lesson for AI Founders: General Purpose AI Will Win

2025-01-12

Lukas Petersson's article highlights a crucial lesson for AI founders: focusing on engineering workarounds for current model limitations is a losing strategy. History shows that general-purpose AI approaches consistently outperform specialized ones. Many current AI products over-emphasize vertical solutions and workflow constraints, neglecting the potential of more capable models. The author argues that focusing on more general, autonomous AI models is key to long-term success in the rapidly evolving AI landscape. The risk is that engineering efforts to improve current models will be made obsolete by future advancements.

Read more