arXiv:2606.15874v1 Announce Type: new Abstract: Every major LLM agent framework gives the LLM the role of orchestrator; the model decides what to do next, when to call tools, and when to stop. We argue that token explosion, control-flow hallucination, and unreliable completion are not implementation bugs but architectural consequences of assigning the deterministic work of looping, branching, and sequencing to a probabilistic system. A better prompt or a stronger model cannot guarantee the reliability of the LLM agent. We therefore propose Agentic Programming, in which the program governs all
The proliferation of LLM agent frameworks has exposed fundamental architectural limitations in assigning deterministic control to probabilistic systems, necessitating a re-evaluation of agentic programming paradigms.
This research proposes a new architectural approach, Agentic Programming, that could fundamentally alter how AI agents are designed and deployed, addressing critical reliability and efficiency issues inherent in current LLM-orchestrated systems.
The proposed Agentic Programming shifts the control-flow from the probabilistic LLM to a programmatic layer, potentially leading to more reliable, efficient, and scalable AI agents.
- · AI agent developers
- · Enterprises adopting AI agents
- · Open-source AI frameworks
- · Software engineering principles in AI
- · Current LLM-centric agent framework designs
- · Companies reliant on brute-force LLM scaling for agent reliability
Increased reliability and performance of AI agents, making them viable for more complex and critical tasks.
A shift in demand towards tools and platforms that support Agentic Programming principles, potentially creating new market leaders.
The development of a new class of AI applications that were previously unfeasible due to the limitations of existing LLM agent architectures.
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