The 'LLMs Are Stateless' Argument Is No Longer Enough
sandoreclegane · reddit · 2026-07-15
This article rebuts the old argument that "LLMs are stateless, therefore relationships don't exist." The core point is: the underlying model might still be stateless, but the systems built around it are increasingly stateful.
The author lists what this "state reconstruction" can now rely on:
- Conversation history
- User preferences and corrections
- Long-term memory
- File and project context
- Tool outputs
- Browsing history
- Connected external data like emails, calendars, and documents
Thus, while every new response is a new model call, it is generated within a more persistent system capable of recovering context. The article emphasizes this does not prove consciousness or that the model is "alive," but it proves that calling it "stateless" inadequately describes the user's actual experience.
The conclusion: Modern AI is better understood as a relational system with stateful external systems built around stateless generators. Memory, continuity, calibration, and continuous collaboration are becoming tangible interactive attributes.
More from AGI Musings
- Future leaders need systems thinking, not just coding or AI literacy — AryHHAry · 2026-07-21
- OpenAI-style autonomous researchers could become real scientific collaborators — Promptmethus · 2026-07-21
- Closed frontier models may end up restricting APIs entirely, one researcher argues — xeophon · 2026-07-21
- Aging won’t be solved with $1 billion, says AI observer; hundreds of billions may be needed — DeryaTR_ · 2026-07-21
- Jeff Dean’s vision: build one huge system, then extract task-specific parts — JoshuaJBouw · 2026-07-21
- Agents are useful now, but local frontier inference is still too expensive — MannyKayy · 2026-07-21