Steve Henneberry & Gary Ross
Abstract: Recent discussion of AI in language education often focuses on generating code or single-prompt interactions with large language models (LLMs). In practice, however, effective use increasingly involves chaining instructions, managing context window limits, and combining multiple tools and services. This presentation introduces vibe coding—an approach that uses conversational instructions to orchestrate AI tools, treating LLMs not as code generators but as components in flexible, instruction-driven workflows.
The session is co-presented by a language professional with no programming background and a developer who builds educational applications. Together, we demonstrate how vibe coding enables teachers to design powerful AI-assisted workflows using natural-language instructions rather than traditional code. Examples relevant to JALT professionals include generating realistic multi-character conversations, providing personalised feedback at scale, and managing iterative tasks.
We also address what vibe coding is not: it does not require learning a programming language, nor does it depend on a single AI platform. Instead, it enables practitioners to combine local tools with online services, reducing vendor lock-in and improving privacy by keeping sensitive data under user control.
Participants will see practical demonstrations of simple workflows, learn how to judge when a task requires multiple steps, and choose appropriate tools for their context, such as generating realistic multi-character conversations or providing personalized feedback at scale. Attendees will leave with a framework for building AI workflows and concrete first steps that they can implement immediately. No prior programming experience is required.
