In 2025, I contributed to the publication of the Japanese translation of
Super Study Guide: Transformers & Large Language Models.
This book provides a comprehensive and systematic overview of how Large Language Models (LLMs) work, covering everything from fundamental concepts to cutting-edge research.
In 2024, I had the opportunity to explore the Googleplex and the Bay View campus in Mountain View.
Biking to Shoreline Park was a surprisingly delightful experience.
My trip also took me to San Francisco, where the Google offices are conveniently located near many popular attractions.
While the Cable Car was a classic San Francisco experience, the Waymo ride was the highlight for me.
Finally, The Stanford University campus was massive and incredibly beautiful.
Walking around was fun, but biking would've been easier to see everything.
Agent Simulation
3D pedestrian simulation game developed for web browsers using Three.js.
The red-shirted agent is controlled by human key inputs or swipes on the touchscreen.
I am attempting to make other agents walk more intelligently.
Tiny Robot
toio cubes seem to be just cute toys but are also full-scale programmable robots.
Simply combining the two example codes
'chase' and
'keyboard-control',
a toio core cube can chase another cube controlled with a keyboard.
The two cubes are connected through Bluetooth to my Mac running Node.js.
Spreading all the additional paper mats sold at
switch-science.com
my toio cubes have 12x space to move around.
One more toio core cube joined.
The first cube is controlled with a keyboard. Other cubes chase the cube in front of them.
You can selfie in front of a nice background with this web app no matter where you are.
Your image will be processed on your device without being sent to remote servers.
TensorFlow.js and
BodyPix
model are used to segment persons from an image.
