This program aims to provide a “third place” for interaction among a wide range of users from academia and industry in the field of computational science, with a particular focus on users of the D3 Center’s large-scale computing systems. In addition to users of the D3 Center, this program is also recommended for people affiliated with academic institutions and private companies that use our systems, as well as those at other academic institutions and private companies who:

• would like to get started with programming using generative AI
• would like to use generative AI for supercomputing tasks, such as creating job scripts
• would like to learn how to choose AI coding tools and understand the risks associated with their use
• already use generative AI but would like to learn how to reduce usage costs

 
This seminar is organized with the cooperation of the D3 Center’s Advanced High-Performance Computing Infrastructure Research Division, led by Professor Date. Graduate students from the division have developed the seminar program and will serve as instructors. Drawing on their own experience, the seminar is intended for those who are considering getting started with “programming using generative AI,” as well as those who have tried it before but gave up and would like to give it another try. Participants are welcome to take this opportunity to freely ask questions about anything they are unsure of.
 

Getting Started with Programming Using Generative AI: The program consists of three two-hour sessions, held once a week over three weeks. In recent years, programming using generative AI (“AI coding”) has rapidly become widespread. Its capabilities now extend beyond simple code completion, and “coding agents” that can autonomously carry out tasks from writing code through execution have reached a practical stage. In supercomputing as well, generative AI is expected to improve efficiency in tasks such as creating job scripts and developing programs.
This seminar consists of three sessions covering the basic mechanisms of AI coding, techniques for using agents effectively, and practical ways to reduce usage costs.

Session 1: Sep. 10 10:00 - 12:00	Introduction Prof. Susumu Date, Advanced High-Performance Computing Infrastructure Research Division, D3 Center, The University of Osaka
	Introduction to AI Coding: Lecture (with Demo) History of AI coding, LLM models and harnesses, connecting to a supercomputer with Cursor, and creating a skill for job execution (Kohei Taniguchi, Advanced High-Performance Computing Infrastructure Research Division, D3 Center, The University of Osaka / Graduate School of Information Science and Technology, The University of Osaka)
Session 2: Sep. 17 10:00 - 12:00	Smarter AI Coding: Lecture Plan mode, subagents, and loop engineering (Kohei Taniguchi, Advanced High-Performance Computing Infrastructure Research Division, D3 Center, The University of Osaka / Graduate School of Information Science and Technology, The University of Osaka)
Session 3: Sep. 24 10:00 - 12:00	Lower-Cost AI Coding: Lecture Token-saving techniques, local LLMs, and free models (such as OpenCode) * The program content is subject to change depending on the availability of models and other circumstances. (Kohei Taniguchi, Advanced High-Performance Computing Infrastructure Research Division, D3 Center, The University of Osaka / Graduate School of Information Science and Technology, The University of Osaka)