maybe you can’t just tack statefulness onto an agent, you have to figure out how to represent the agent as a state

This post explores the architectural challenge of implementing state management in AI agents, arguing that statefulness cannot simply be added as an afterthought but requires fundamental redesign of how agents are represented internally. Nakajima suggests that proper agent state representation needs to be integrated into the core system design rather than treated as a peripheral feature.

Source: Yohei Nakajima (X) | 2026-05-18