Beyond "Anyone Can Build": Social Infrastructure and Felt Legitimacy in Vibe Coding Hackathons

Abstract

Vibe coding is often described as lowering the technical barrier to software development by allowing people to build through natural-language interaction with AI code-generation tools. However, less is known about the social conditions through which people come to feel able and legitimate to participate in AI-mediated software development, especially in hackathon settings. Drawing on research on AI-mediated software development and hackathons as participatory design, I ask what social conditions matter when software development is no longer limited to people with conventional programming roles. I report nine semi-structured interviews and anonymized fieldnotes from three in-person vibe coding hackathons. Interviewees valued watching others build, asking questions during facilitated sessions, receiving feedback, contributing non-coding expertise, and judging whether AI-generated software was "good enough" to show. I frame these events as social infrastructure for felt legitimacy, where facilitation, peer visibility, demos, and feedback help participants move from "AI can generate software" to "I can legitimately contribute to software development with AI." The paper offers an early empirical and conceptual framing of vibe coding hackathons, showing why HCI should study the social arrangements that make AI-mediated software development feel legitimate and possible.