Tokyo AI (TAI) retweeted
Aidan Gomez is coming to @TokyoAI_JP for the first time. He co-wrote “Attention Is All You Need,” the Transformer paper behind modern LLMs, and co-founded @cohere.
Join us Nov 12 at 6pm for a fireside chat with Aidan and demos from Cohere’s engineers.
luma.com/ns8bpuiz
Tokyo AI (TAI) retweeted
New @TokyoAI_JP event on open-endedness - Thurs 12th Nov @ 6pm 🗼
We'll be joined by @miz_oka and @BertChakovsky to discuss how artificial systems could keep producing novel ideas or behaviours beyond what they already know 🦠🐛🦏
Tokyo AI (TAI) retweeted
Applications now OPEN for the Tokyo AI | OpenAI 100-Hour Game Builder Challenge. Apply until Sept 10 or when spots are filled. >$100K in OpenAI prizes up for grabs. Judges include gaming, AI, and anime pioneers and leaders. More on LUMA.
👉 luma.com/3kj24doy
Tokyo AI (TAI) retweeted
I'm speaking at @TokyoAI_JP next week about agentic memory for edge devices, sharing early results from @SakanaAILabs on long-running agents and lessons on context eviction, retrieval, and which past information should shape the next action.
Tokyo, Sept 4th 19:00: luma.com/5n2lg7p7
Tokyo AI (TAI) retweeted
Amazing turnout to hear @ciaran_regan_ talk about @SakanaAILabs' Continuous Thought Machines 🌀
I'll be presenting our work on Continuous Thought Machines at the @MistralAI x @SakanaAILabs x @huggingface meetup in Tokyo today!
Tokyo AI (TAI) retweeted
/
「Engineering Agentic Workflows: A Full-Stack Perspective」が開催中です!🌍🛠️
\
Tokyo AI主催の本イベントがスタートし、エージェント型AIシステムの実装に関する実践的な議論が進んでいます✨
会場は満席となり、多くの参加者が集まる中、エンジニア視点での知見共有が行われています💻
長時間推論ループやエージェントスキルの設計、プロトタイプから本番運用へとつなげるアプローチなど、現場で活用できる内容が紹介されています。
この後のセッションや交流も楽しみです🤝
==========
/
“Engineering Agentic Workflows: A Full-Stack Perspective” is now underway! 🌍🛠️
\
Hosted by Tokyo AI, the event has kicked off with practical discussions on building and deploying agentic AI systems ✨
The venue is at full capacity, with many participants gathered and engaging in knowledge sharing from an engineering perspective 💻
Topics include long-running inference loops, structured agent skills, and approaches for moving from prototypes to production-ready systems.
We’re looking forward to the sessions ahead and the networking to follow 🤝✨
#SAKURADEEPTECHSHIBUYA #SDS #サクラディープテックシブヤ #AI
Tokyo AI (TAI) retweeted
Kicking off @TokyoAI_JP "Scaling Physical AI" with @lucanunz on general purpose embodied AI for assistive robotics
Tokyo AI (TAI) retweeted
(De)-Vibe Coding Hackathon
A rising Silicon Valley startup, @ccdotdev, is bringing the De-Vibed Hackathon to Tokyo.
-> Code quality over vibe coding.
💰 $5,000 prize
🤝 Co-hosted with @TokyoAI_JP & @OpenDataCircle
📍 In-person @ LODGE, LY Corp HQ (Tokyo)
Register: luma.com/efebvv48
Tokyo AI (TAI) retweeted
🚀REWW-ARM (Remote Wire-Driven Mobile Robot)に関する論文がAdvanced Intelligent Systemsから出版されました!
一切の電気系を含まない, 水の中も動ける, 長距離移動可能な蛇型のワイヤ駆動ロボットです!
どのように効率的にワイヤ張力を伝達するかが肝になっています!
hatofly.github.io/reww-arm-r…
🚀Our latest paper is now published in Advanced Intelligent Systems!
""REWW-ARM -- Remote Wire-Driven Mobile Robot: Design, Control, and Experimental Validation""
📄Paper ↓
advanced.onlinelibrary.wiley…
Tokyo AI (TAI) retweeted
I was so grateful to be able to attend #NeurIPS2025 in San Diego. I wrote a retrospective from a metascientific perspective.
medium.com/@rmaruy3/metascie…
TL;DR: NeurIPS is a frontier of huge shift in science by AI. Despite that, for now, NeurIPS remained very human.
Tokyo AI (TAI) retweeted
AI x Robot hackathon with @AMD Japan kicking off.
Tokyo AI (TAI) retweeted
本日19:00より、@KERNEL_DEEPCORE さん・@TokyoAI_JP さんと合同で
第2回 Physical AI 勉強会 を開催します!
当日参加も大歓迎です!参加費無料&軽食つきなので、気軽にお立ち寄りください!
📅 11/28(金) 19:00–21:00
📍 KERNEL HONGO
🔗 connpass.com/event/373516/
#PhysicalAI勉強会 #PhysicalAI
Tokyo AI (TAI) retweeted
An example of AI-humans coexistance, from Japan: bowing to a car to let it pass (with an apology for the road works)?
I was impressed when my team showed me this example of intelligence and behavioural generalisation.
Watch this — seeing a person bow to welcome our car to move forward is definitely not something I've seen outside of Japan. Yet in this clip you can see our AI Driver recognise that this person is asking us to stop. We have the intelligence to understand the context that we shouldn't nudge around them, rather hold in place, despite no signage or red light. Then, if you keep watching, you'll see we recognise the bow and a wave as permission to drive through the road works.
Less than 1% of our training data is from Japan. Our AI is trained end-to-end. There are no explicit rules or labelling, no explicit gesture or body pose recognition, this is all learned capability that is scalable and data-driven.
This is how @wayve_ai will scale autonomous driving to the world and address the long-tail of edge cases.
Tokyo AI (TAI) retweeted
Living and building AI in Tokyo is bringing lots of satisfaction lately. And I feel like we are just starting!
Genki Robotics, a new humanoid robotics startup, is headquartered in Tokyo, Japan.
It's founded by Andy Rubin, the founder of Android, who was a Google executive for nine years. Rubin led Google’s robotics division during 2013–14, which included Boston Dynamics, which Google had acquired. Andy Rubin’s investment firm invested in Agility Robotics in 2018.
Genki Robotics's vision "is to accelerate the integration of intelligent humanoids into society — augmenting human capability and shaping the future of work and service." The company is backed by Andreessen Horowitz and AMD, among others.
Here's Andy Rubin in 2018 talking about why the bipedal form makes sense, with Agility's Cassie robot on stage
Tokyo AI (TAI) retweeted
What better place to discuss robots and society than @TokyoAI_JP?
First is @aopico (@yukaikk) on designing robots that feel and can make emotional connections with humans 🤗
Tokyo AI (TAI) retweeted
Tonight’s session by @TokyoAI_JP covered Japan’s healthtech and medtech ecosystems, with @toluwajosh sharing insights from his experience working in the field.
🇯🇵 Key points on market trends and regulation in Japan:
👵 Demographics is the driver
About 30% of Japan’s population is over 65. Senior care remains the main area of near-term demand. Private clinics tend to adopt new technologies faster than larger public hospitals, offering initial entry points for new products.
⚙️ Policy direction
National initiatives under Society 5.0 are creating opportunities for controlled testing environments. Shibuya’s special designation, for instance, allows pilot projects in areas such as resident monitoring and digital services.
💸 Economic context
Many hospitals operate under financial pressure due to the national fee schedule. Solutions that reduce administrative work, streamline patient flow, or prevent repeat visits are gaining attention if they can show measurable operational benefit.
🤖 Current AI applications
Adoption is growing in imaging support, scheduling, pharmacy coordination, and early digital therapeutics. In this environment, reliability and integration into clinical workflows are more important than novelty.
📜 Regulatory framework (PMDA)
Japan has defined pathways for Software as a Medical Device (SaMD), with classifications based on risk. Reviews can move quickly once documentation is complete, translated into Japanese, and supported by a local Marketing Authorization Holder (MAH). A conditional approval process introduced in 2023 allows real-world data collection before full approval.
🚀 Expedited options
Priority review and the Sakigake designation can shorten approval times for innovations with clear clinical relevance, particularly if developed with a Japan-first approach.
🤝 Go-to-market approaches
Effective entry strategies often involve collaboration with universities or key opinion leaders, partnerships with established distributors, and early pilots that can produce measurable data. Trust and local relationships are important factors.
📈 Adoption criteria
In a cost-sensitive healthcare system, new solutions must show clear advantages over the standard of care, focusing on higher diagnostic accuracy, improved outcomes, or time savings for clinicians.
In brief, Japan’s healthcare sector values clinical rigor and demonstrable benefit. Companies that provide strong evidence, localize effectively, and build the right partnerships can find practical pathways to adoption.
Please reach out to Joshua Owoyemi or to us for more info and connections in the Japanese Medtech and Healthtech ecosystem.
Thanks to @ITAtradeagency - Tokyo for the patronage!
Tokyo AI (TAI) retweeted
Mistral AI Meetup in Tokyo #MistralMeetup
Mistral AI の東京初イベント!Hugging Face x Tokyo AI!私はお手伝いをします!よかったらいらしてください!
aimeetup.connpass.com/event/…
Tokyo AI (TAI) retweeted
Sliding window attention (SWA) is powering frontier hybrid models for efficiency. Is there something better?
Introducing Phalanx, a faster and better quality drop-in replacement for sliding window attention (SWA).
Phalanx is a new family of hardware and numerics-aware windowed layers designed with a focus on data locality and jagged, block-aligned windows that map directly to GPUs.
In training, Phalanx delivers 10–40% higher end-to-end throughput at 4K–32K context lengths over optimized SWA-hybrids and Transformers by reducing costly inter-warp communication.
Today, we are releasing both the technical report, a blog, and Phalanx kernels in spear, our research kernel library.
We are hiring.
Tokyo AI (TAI) retweeted
Join us tomorrow for an in person session organized by
@TokyoAI_JP @ikulyatin @__MLT__ and myself.
“How do robots see, imagine, and act?”
🔹 Seeing - VLMs
🔹 Imagining - World Models
🔹 Doing - VLAs
👉 Register: luma.com/9wwhomfv






























