Agentic systems with artificial intelligence: transforming software development and its application in agriculture

Abstract

Extended version of the agentic systems talk, aimed at AGROSAVIA. It opens with search trends for agentic AI, Stanford’s 2025 AI Index Report, AI in Latin America, AI investment and compute capacity (GPUs, TPUs) and tools like JAX, PyTorch, TensorFlow and Keras 3. It reviews neural network architectures, Transformers, the ChatGPT moment, advances such as Chain-of-Thought, RLHF and ReAct, and reasoning models. It defines agents and presents ecosystems (LangChain/LangGraph, n8n, Google ADK, Agno, Unsloth), protocols (MCP, A2A, AG-UI), context engineering and coding assistants (Claude Code, Gemini CLI, Windsurf, Cursor, NotebookLM). It shows applied cases such as Cereza, Porkus, Nutribot and Feedibot.