Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
Fugu Model in the Physical World: Multi-Agent AI with Reachy Mini and AxiDraw
See Fugu AI orchestrate robotics with Reachy Mini and AxiDraw. Witness multi-agent reasoning and planning visualized on paper in this live demo.
A live demo of Fugu coordinating AI workflows through two physical interfaces: Reachy Mini as an embodied conductor for multi-agent reasoning, and AxiDraw as a pen-plotter that turns Fugu’s planning, correction, and orchestration traces into visible paper artifacts.
Sakana Fugu orchestrates multi-agent frontier models via APIs for reasoning.
- FuguProject Fugu is a collaborative Chromium initiative that equips web applications with powerful, native-level operating system capabilities.Led by Google, Microsoft, and Intel, Project Fugu bridges the gap between web and native applications by introducing advanced APIs directly to the browser. This initiative enables web developers to safely access low-level hardware and system features (such as the File System Access API, Web Bluetooth, and Web USB) while preserving the web's core security model. By standardizing these capabilities, Fugu allows Progressive Web Apps to deliver deep, desktop-class experiences without the friction of platform-specific app stores.
- Reachy MiniDeploy embodied AI from your desk: Reachy Mini is the open-source, 11-inch desktop robot (6 DoF head, Python SDK) built for human-robot interaction and rapid prototyping.This is the Reachy Mini: the first open-source desktop robot for embodied AI and Human-Robot Interaction (HRI). Developed by Pollen Robotics and Hugging Face, this 11-inch, 3.3 lb unit is a powerful, accessible platform starting at $299. It features a 6 DoF head, full body rotation, and a comprehensive Python SDK for immediate coding. Developers integrate seamlessly with the Hugging Face Hub, accessing over 1.7 million AI models for vision and speech tasks. Choose the Lite (wired) or Wireless (onboard Raspberry Pi 5) model: both deliver a robust environment for prototyping and sharing real-world AI applications directly from your desk.
- AxiDrawA high-precision personal writing and drawing robot that uses real pens to produce authentic, handmade-quality vector art and handwriting.AxiDraw is a versatile pen plotter designed to bridge the gap between digital design and physical paper. By holding almost any standard pen (including fountain pens, permanent markers, and chalk crayons), it translates vector graphics from software like Inkscape into precise, continuous lines. The machine excels at tasks requiring an authentic human touch, such as signing diplomas, addressing formal invitations, or creating intricate generative art. Operating with a quiet, open-ended design, it handles paper of any size (even extending beyond the machine itself) to deliver flawless execution without the sterile look of standard inkjet printing.
- sakanaTokyo-based Sakana AI builds next-generation foundation models inspired by nature, evolution, and collective intelligence.Founded in Tokyo by former Google researchers David Ha and Llion Jones (co-author of the landmark Transformer paper), Sakana AI takes a unique approach to artificial intelligence by mimicking natural systems. Instead of building monolithic systems, the team focuses on evolutionary optimization and collective intelligence (similar to a school of fish, or "sakana" in Japanese). Their notable breakthroughs include "The AI Scientist" (an automated system that conducts independent machine learning research) and pioneering methods to automatically merge the knowledge of multiple large language models without massive computational overhead.
- beta accessBeta access bridges the gap between pre-release software and real-world validation by putting unfinished products into the hands of early adopters.Before a digital product hits wide release, it needs real-world stress testing to survive. Beta access provides this environment by distributing pre-release software to a targeted group of external users (whether through closed, invite-only cohorts or public, open-enrollment pools). Product teams use feature flags and dedicated staging platforms to deploy these early versions, allowing them to monitor system performance, identify critical bugs, and gather usability feedback. This structured phase mitigates launch risk, ensures the final release meets user expectations, and builds an active community of early adopters before the general public ever logs in.
- HTTPSHTTPS secures web traffic by wrapping standard HTTP in a TLS-encrypted tunnel to ensure data integrity and privacy.HTTPS (Hypertext Transfer Protocol Secure) is the industry standard for safe data exchange, utilizing Transport Layer Security (TLS) to encrypt communications between browsers and servers. By leveraging Port 443 and X.509 digital certificates, it prevents man-in-the-middle attacks and verifies server identity. Modern browsers like Chrome and Firefox now enforce HTTPS by default, flagging non-encrypted sites as insecure to protect sensitive user credentials and financial transactions.
Related talks
More from the community
TokyoTaiwanRadar: Multi-agent event platform
Tokyo
Explore Tokyo Taiwan Radar, a multi-agent platform for Taiwan-Japan exchange. See system, code, and architecture, focusing on dynamic…
Debugging your AI agent in local
Tokyo
This talk explores why debugging AI agents locally is difficult, detailing why traditional methods fail and presenting effective…
OpenClaw-Powered Port: Turning a Figma Plugin into a Standalone Web App with Claude + Grok
Tokyo
See how AI agents transformed a Figma plugin into a viral web app, demonstrating practical AI orchestration for…
SafeGuide: An Offline AI Assistant for Emergency Guidance
Tokyo
Learn how SafeGuide runs a compact language model locally to deliver clear, actionable safety instructions during emergencies without…
Design-by-Transparency: Fixing Authority Before Execution
Tokyo
Learn how to fix AI authority before execution. This talk demonstrates explicitly defining intent, authority, and stop conditions…
How we made our RAG truly multimodal
Tokyo
A deep dive into building multimodal RAG from scratch, covering data ingestion, embedding comparisons (CLIP vs. others), and…
Compose Email
Loading recent emails...