Micron Begins ¥1.5 Trillion Fab Expansion in Hiroshima — Toward Mass Production of HBM for AI

On July 4, Micron Technology held a groundbreaking ceremony for an expansion of its plant in Hiroshima Prefecture. The company will invest a total of ¥1.5 trillion (approximately $9.3 billion) to strengthen its manufacturing base for high-bandwidth memory (HBM), which is essential for AI processors such as those from Nvidia. Shipments are expected to begin around summer 2028.

Japan's Ministry of Economy, Trade and Industry (METI) has allocated up to ¥500 billion in support, signaling its commitment to strengthening the semiconductor supply chain and building a domestic production base for AI infrastructure. With AI demand set to keep growing and stable memory chip supply becoming a critical issue, large-scale investments in Japan are continuing.

China Issues New Guidelines Encouraging AI Use in Movie Theaters

China's National Film Administration and the State Administration for Market Regulation have released new guidelines urging movie theaters to go beyond traditional film screenings and offer a wider range of services.

The guidelines recommend deploying AI agents, operating karaoke and cafés, selling film-related merchandise, and hosting art exhibitions and pop-up stores. Theaters are also encouraged to convert lobbies and unused auditoriums into cultural event and retail spaces — a push toward turning cinemas into multi-experience complexes.

With China's film industry facing declining audiences and the rise of streaming, the move amounts to policy-level backing for the search for new AI-powered revenue models.

"AI Wrote the App — But Who Fills the Security Holes?"

A technical article published on Qiita takes a sharp look at the question of security responsibility for AI-generated code. It raises a fundamental issue: when vulnerabilities lurk in applications built with AI tools, who should bear responsibility — the developer, the company, or the AI vendor?

As AI code generation becomes an everyday practice, building a "working app" has never been easier — but working out how to build security review into the process has become urgent. The article discusses blind spots unique to AI-generated code and directions for practical countermeasures.

This is a topic that will only grow in importance as AI coding tools proliferate.

An Implementation Log: Building a Real-Time Voice Avatar with the Gemini Live API

A developer at クリスタルメソッド株式会社 has published an implementation write-up on Qiita detailing a virtual human (conversational AI avatar) built with Google's Gemini Live API (bidirectional WebSocket voice interaction).

Particularly interesting is the design that uses Function Calling to control the avatar's nodding. A mechanism that reproduces natural back-channel responses timed to the flow of spoken dialogue is a key element in enhancing the expressiveness of real-time AI avatars.

The post also includes verification of response quality by profession, quantitatively analyzing how AI responses differ across use cases.

Results Announced for "Unslop," an AI-Written Fiction Contest

The results of the 2026 Unslop AI-Written Fiction Contest, a competition for fiction written with AI, have been announced. The initiative honors high-quality human-AI co-created works that are not "slop" (shoddily mass-produced AI output), and it has been drawing attention on Hacker News.

The contest positions AI not as a mere auto-generation tool but as a partner that extends human creativity, and it is drawing attention as an example of what creative work can look like in the AI era.

---