The field of artificial general intelligence (AGI) is moving towards exploring the possibilities of general intelligence using large language models (LLMs). Researchers are identifying and understanding specific mechanisms and representations sufficient for general intelligence, with a focus on architectures and cognitive design patterns. The use of LLMs is offering a new combination of mechanism and representation for exploring general intelligence, particularly in reasoning and interactive use cases. Noteworthy papers in this area include:

• Architectural Precedents for General Agents using Large Language Models, which summarizes recurring cognitive design patterns in various pre-transformer AI architectures and explores their application in systems using LLMs.
• Boosting Performance on ARC is a Matter of Perspective, which achieves state-of-the-art performance on the Abstraction and Reasoning Corpus (ARC-AGI) using task-specific data augmentations and a depth-first search algorithm.
• AI Agents vs. Agentic AI: A Conceptual Taxonomy, Applications and Challenge, which critically distinguishes between AI Agents and Agentic AI, offering a structured conceptual taxonomy and challenge analysis to clarify their divergent design philosophies and capabilities.