【Jカーブはもう古い。】AI時代の起業は「量産」が正解？連続でプロダクトを生むAIdeaLabの戦略とは？AIdeaLab冨平×田所 前編

AIdeaLab CEO Tomihira Shun explains how his startup studio mass-produces AI products using vibe coding and a ‘Mu-curve’ strategy that replaces the traditional J-curve.

• AIdeaLab has shipped Japan’s first image generation app (AI Picasso, 3rd–4th globally at Stable Diffusion launch), a real-time AI clone of YouTuber Hiroyuki, and AI Terapo-kun outbound call agent.
• Comment Screen, Tomihira’s pre-AIdeaLab product, hit 200,000 users in one month after Japan’s COVID state of emergency; it was adopted by over half of Japanese universities.
• AI Terapo-kun generated controversy (炎上) but accelerated growth; it books real sales appointments and callers rarely detect they are speaking to an AI.
• AIdeaLab’s moat thesis for AI Terapo-kun: outbound call-center market is large, big tech avoids it due to regulatory and reputational risk, and appointment-rate data provides a clean optimization label.
• Tadokoro argues AI has compressed the J-curve into a ‘Mu-curve’: AI lowers build cost so founders can mass-release multiple product variants and run real-time AB tests across customer segments simultaneously.
• Tomihira uses vibe coding to ship MVPs in roughly one week, iterating on prompt/talk-script rather than code after launch.
• AIdeaLab is developing a proprietary text-to-video and animation generation model it plans to release as a platform.

2026-01-21 · Watch on YouTube