Coding with AI and Vibe Coding

Large Language Models (LLMs) have revolutionized coding tasks by leveraging extensive training data and advanced next-token prediction capabilities. This presentation explores the diverse landscape of coding models, their core functionalities, training methodologies, and the state-of-the-art (SOTA) systems driving current advancements. It also examines the key benchmarks used to evaluate model performance. Additionally, it looks into the growing trend of "vibe coding"—a practice embraced by both professional and non-professional developers—highlighting its best practices, inherent risks, and limitations. Finally, the discussion covers the tools and platforms that support vibe coders, offering insights into how this evolving paradigm is reshaping software development workflows.