arXiv:2606.30306v1 Announce Type: cross Abstract: Always-on agents are systems whose future behavior depends on durable state accumulated across earlier interactions. We treat them as persistent-state systems: the operative system includes retrievable memories, but also task ledgers, permissions, credentials, commitments, provenance and audit records, shared state, trigger conditions, and externally committed effects linked to those records. The survey reads the literature through six diagnostic axes for each state item, authority, scope, mutability, provenance, recoverability, and actionabili
The proliferation of LLMs is forcing a direct confrontation with the challenges of managing persistent, stateful, and context-rich AI agents, making this survey timely.
This survey provides a foundational framework for understanding and building future AI agents, moving beyond stateless models to autonomous entities with memory, ethics, and accountability.
The focus shifts from isolated LLM interactions to architecting 'always-on' intelligent systems capable of sustained, multi-interaction reasoning and action, impacting how AI-driven value is created.
- · AI platform developers
- · Enterprise software vendors
- · Cybersecurity firms
- · Legal and compliance tech
- · Companies relying on stateless AI
- · SaaS providers resistant to agent integration
- · Human-in-the-loop 'orchestration' services
- · Simple workflow automation companies
Architects and developers will prioritise persistent memory, state management, and governance features in their agentic systems.
New standards and regulatory frameworks will emerge for agent ethics, accountability, and data provenance as these systems become more autonomous.
The definition of 'agency' in AI will expand beyond simple task execution to encompass long-term commitment, shared state, and auditable history, profoundly reshaping white-collar work and service industries.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI