The Who's Who of AI

“日経新聞による、グーグルディープマインド東京を率いる全炳河（Heiga Zen）氏の素晴らしい特集記事。2018年、Heigaさんと2人でGoogle Brain東京チームを立ち上げた日々を懐かしく思います。当時は時差の厳しい深夜の会議をこなしながら、日本のAI研究の存在感を示すために必死でした。現在、彼がGDM Tokyoを率い、私が Sakana AI を起業して、東京のAIエコシステムがここまで大きく成長したことを本当に嬉しく思います！ www.nikkei.com/article/DGXZ...”

“Deep learning relies on backprop, but the brain can't implement it exactly. How can physical systems solve multilayer credit assignment? We draw on distributed optimization and NeuroAI, where predictive coding uses local error dynamics, but credit signals s…”

“By orchestrating the world's models, we are building the resilient infrastructure required for AI sovereignty. Try: sakana.ai/fugu Blog: sakana.ai/fugu-max-rel... 🐡”

“Crucially, Fugu Ultra v2 does all of this without Fable 5, Fable 5.1, or GPT-6-Astra in its agent pool. The orchestration system evolved with model resiliency in mind. It does not rely on individual frontier models to deliver frontier output. sakana.ai/fugu”

“Fugu Ultra v2 is now live on OpenRouter! 🐡 openrouter.ai/sakana/fugu-... Our flagship orchestration engine built for peak performance on complex multi-step reasoning, autonomous research, and full-stack software development.”

“Fugu Max is live on Open Router! 🐡 Our learned multi-agent orchestration engine routes tasks across open-weights and specialized models for Pareto-frontier efficiency. • $2/$6 per 1M tokens • Multimodal (Image/PDF) + Web Search • Configurable reasoning & Fu…”

“PC-ALM propagates credit to arbitrary depth, even in deep narrow nets where standard PC fails. We want to understand how the brain computes credit locally, and aid neuromorphic hardware where dynamics are cheaper than GPUs. Paper: arxiv.org/abs/2605.31022 C…”

“Trained detectors fail when generators or styles change. Our Percept-Lens benchmark tracks these shifts. When they fail, is the model blind to the difference, or just the decision rule? arxiv.org/abs/2608.18523”

“• Smart Cellular Bricks: How can physical systems achieve collective intelligence and self-repair without a central brain? We hope that this encourages other researchers to also explore different paradigms, and take a leap of faith with us. “Greatness canno…”

“The latest issue of Scientific American features our Smart Cellular Bricks research: simple cubes that collectively recognize their own shape and repair themselves without a central brain. Thread 🧵”

“From simulation to 197 physical bricks with 100% accuracy, this is a step toward machines that build and heal themselves. Our blog post: sakana.ai/smart-cellul... Nature Communications paper: www.doi.org/10.1038/s414...”

“From simulation to 197 physical bricks with 100% accuracy, this is a step toward machines that build and heal themselves. Our blog post: sakana.ai/smart-cellul... Nature Communications paper: www.doi.org/10.1038/s414...”

“We froze a general vision model, fit a simple Gaussian to its features, and outperformed the strongest released detector. The real-vs-generated signal is already in general features. Better models help, but better rules matter too. pub.sakana.ai/percept-lens/”