Jean Paniagua, a Software Engineer at Gorilla Logic, recently delivered a lecture on the foundations of Artificial Intelligence (AI) to second- and third-year Business Informatics students at the University of Costa Rica (UCR) in Tacares.
The session, titled “Artificial Intelligence: Foundations, Generative AI & Prompt Engineering,” provided students with a comprehensive overview of how modern AI tools are reshaping the technology landscape. Drawing from five years of experience as a web developer, along with two years specifically coding with AI acceleration, Paniagua bridged the gap between academic theory and real-world application.
Demystifying the Artificial Intelligence “Magic”
A core theme of the lecture was the importance of dispelling common myths surrounding AI. Paniagua explained that AI is neither “magic” nor “conscious.” Instead, he described it as “powerful pattern matching at scale.” He clarified that while tools like ChatGPT and GitHub Copilot can generate fluent, human-like responses, they do not possess emotions, intent, or true understanding.
Students learned that modern AI operates by predicting the next most likely piece of information based on massive datasets. Paniagua summarized this idea succinctly: “AI predicts; it doesn’t think.”
From Copilots to Autopilots
The lecture guided students through the rapid evolution of AI systems. Paniagua outlined the progression from traditional machine learning to today’s Generative AI, and then to the emerging field of “Agentic AI.”
Though current tools act as reactive assistants (Copilots), the future will see autonomous agents capable of planning tasks, executing multi-step workflows, and even correcting their own errors. Paniagua compared this shift to moving from the role of a “Junior Developer” to that of an “Autonomous Team Member.”
The Art of Prompt Engineering
A significant part of the talk focused on prompt engineering, the skill of crafting effective instructions for AI models. Paniagua described this process as “programming with language,” noting that the quality of AI output depends directly on the user’s input.
He shared practical frameworks for creating strong prompts, advising students to include specific roles, context, tasks, constraints, and output formats to achieve optimal results. He also highlighted how developers use these techniques to explain unfamiliar code, generate tests, and refactor legacy systems.
Artificial Intelligence: A Tool, Not a Crutch
The session ended with a critical message about responsibility. Paniagua urged students to view AI as a powerful tool, such as a tutor or pair programmer, rather than a shortcut for exams or a blind source of truth. Since AI can sometimes “hallucinate” or provide confidently wrong answers, he emphasized the need for vigilant human oversight.
“Engineers who understand AI will outperform those who don’t,” Paniagua stated in his final remarks, reminding future professionals that verifying AI output is just as important as reviewing a junior colleague’s code.
The lecture offered UCR students valuable insights into their future careers, providing them with knowledge to harness AI responsibly as they prepare to enter the workforce.