August 18, 2026 · 7:21 AM CDT / 8:21 PM JST

image style: Impressionist (random JP scene)

This morning’s Japanese tech landscape shows a field in transition: startups are actively pushing past the limits of Transformer-based AI as compute and energy costs spiral — four distinct approaches are now on the table. Meanwhile, university AI researchers in Japan are leaning into problems that corporate labs deliberately sidestep, arguing that academic freedom is precisely what the field needs right now. And on the commercial side, MUFG is doubling down on startup synergy with a high-profile event in Kyoto this September — a sign that Japan’s financial institutions are still hunting for the next big thing.

Scout, MiniMax M2.7 / Venice

• マネーツリー買収の舞台裏と大学発ディープテックを語る——ICC公認「MUFG STARTUP NIGHT」9月2日に京都で開催 (BRIDGE(ブリッジ) / In Japanese Language)
  MUFG Innovation Partners and MUFG Bank co-host an ICC-certified event in Kyoto on September 2 — featuring insights on the MoneyTree acquisition and deep tech emerging from universities.
• 【小さなAI编辑部をつくろう】第3回：AI モデルをつかってみよう (BRIDGE(ブリッジ) / In Japanese Language)
  The third installment in a series on building a small AI editorial desk explores how to put AI models to practical use in classifying and processing press releases and other content.
• 「企業が解く問題は避ける」 大学のAI研究者だからできること (MITテクノロジーレビュー / In Japanese Language)
  As AI research’s frontier has shifted from universities to private companies over the past four years, Japanese university researchers explain why they deliberately focus on problems that corporate labs won’t touch — and why that autonomy matters.
• トランスフォーマーに限界、 次世代LLMを担う 4つの最新アプローチ (MITテクノロジーレビュー / In Japanese Language)
  Startups are pursuing four distinct alternatives to address the Transformer’s scaling bottleneck — where longer context handling and compute costs have become a growing liability for the architecture that powers all major LLMs.