A J-pop idol ships a full livestream stack—without writing code

Karin Miyamoto, the former Juice=Juice member now pursuing a solo career, has ignited Japan’s tech-and-pop culture conversation by publishing a two-part blog on August 1 detailing how she built an entire system to run a 10-hour YouTube livestream—using Anthropic’s AI coding tool, Claude Code—without writing a single line of code herself. The marathon stream, staged on July 31 to celebrate her 5th single “HANAKIN/Shanikamaa,” turned her living room into a DIY music-video studio and the internet into her control room. The posts—an overview and a technical breakdown—have drawn praise on X (formerly Twitter), where engineers and fans alike called her planning and decision-making “spot on” and marveled that an idol would publish a full “tech blog.”

A participatory concept rooted in Japan’s fan culture

Miyamoto’s idea tapped into Japan’s uniquely engaged fan communities and enduring CD culture. Viewers who bought the CD and posted with a designated hashtag on X helped “unlock” items she could use live—makeup, outfits, even a video-editing app—progressively enabling her to recreate the official music video at home. This playful, participatory mechanic not only rewarded supporters in real time but showcased the polished showmanship and meticulous planning long associated with Japanese pop acts. It also offered an accessible blueprint for creators worldwide: blend audience actions with live production to make fandom tangible.

Four systems, one show: what she built with AI

According to her blog, Miyamoto assembled four key components: a real-time on-screen gauge that counted CD purchases and support posts; a mobile-friendly admin console to adjust numbers and captions from a phone; a hashtag aggregation mechanism that tracked X posts; and an AI-based scoring tool that compared her at-home recreation with the original MV across categories such as composition, costume, and lighting. She instructed Claude Code in Japanese—“I want a screen like this,” “Please add this function”—and, when errors surfaced, pasted the error messages back to the AI for fixes. She emphasizes she neither read nor edited the code, focusing instead on direction and verification.

Cloud where it counts, local when it’s heavy

Her technical article outlines a pragmatic architecture: state that needed to be shared across devices lived in the cloud, while compute-heavy video analysis ran locally. She used Cloudflare Workers and Workers KV to manage the gauge and post counts across multiple devices and accounts, ensuring consistency on stream. The AI scoring for video—large files and heavier processing—ran on her own PC. Anticipating the one-shot stakes of live broadcasting, she layered in manual overrides so numbers could be updated even if automation failed. She initially planned to call Anthropic’s API but, after card payment troubles, switched to OpenAI’s API. Total build time: 10 days.

Why this matters: Japan’s creative tech moment

Reactions from engineers online underscore that this wasn’t merely “AI wrote some code.” They praised the clarity of her requirements, sensible trade-offs, and real-world safeguards—hallmarks of solid systems thinking. For international readers, Miyamoto’s project crystallizes a broader Japan trend: mainstream creators comfortably blending pop culture with practical engineering, an echo of the nation’s monozukuri ethos—crafting with precision and care. As AI lowers barriers to software development, Japan’s artists, indie producers, and small teams can prototype ambitious, audience-centric experiences without large engineering budgets. It’s an encouraging sign for startups, labels, and solo creators eyeing Japan as a launchpad for innovative, global-ready content.

The takeaway, in her own words

Miyamoto concludes that you don’t need to write code if you can clearly express what you want and how it should behave—AI can help you get remarkably far. Her “instruct and verify” method, combined with conservative fail-safes, delivered a resilient production for a high-pressure, 10-hour live performance. It’s a compelling case study for creators in and beyond Japan: define the vision, communicate precisely, and let AI turn ideas into working systems—while you focus on showmanship and audience delight.