In 2023, there were notable limitations of Large Language Models (LLMs), including content hallucinations, small query context, and prompt engineering magic. However, several developments have addressed these issues. RAG (Retrieval Augmented Generation) systems have become a prominent technique for grounding LLMs in external information, reducing the "Content Hallucination" problem. Efficiently connecting LLMs to tools has also improved, with developer tools like LangChain and LlamaIndex gaining popularity. Larger context sizes have enabled models like Google's Gemini 1.5, which can handle near-infinite memory capacities. The LLM ecosystem is maturing, leading to cost reductions, with open-source libraries making it easier for developers to experiment with LLMs. Fine-tuning has evolved to focus on increasing model capabilities in specialized domains, and AI alignment actions are being initiated by governments and companies to ensure that LLM growth is regulated safely and ethically. The mindset shift towards flow engineering, which involves an interactive and step-wise procedure to improve AI reasoning, is expected to be a key development area in 2024. This paradigm shift will enable the construction of iterative step-by-step flows for various AI-related tasks, making it easier to further AI reasoning.