origin
The kid who wanted a Pokédex.
I started coding at 12 to build a Pokédex: a first attempt to make a system personal, useful, and worth understanding.
The kid who wanted a Pokédex. I started coding at 12 to build one, then spent years jailbreaking phones, rooting devices, and customizing PCs. I was drawn less to novelty than to the question underneath: how does this system work, and how could it become more personal and useful?
Learning by building. Smart Manufacturing gave me a physical-systems foundation; HackByte made technical community a responsibility; Project EKA and Qwen3 MoE took the questions into models; Atlan exposed the reality of agents in production; and Waldo is where I am testing what durable, user-owned agency can mean.
Read what shapes the work →01 / Principles
A working compass
Keep meaningful authority with the person.
An agent can prepare, recommend, and act within permission. Remembering more never silently grants it more authority.
Do not confuse a finished run with a completed outcome.
An artifact, evidence, acceptance, and the remaining open loop are different kinds of truth.
Make personal context correctable and user-owned.
Useful memory needs source, time, scope, correction, expiry, revocation, and deletion.
Make powerful systems inspectable.
The more a system can do, the easier it should be to understand what happened and intervene.
Let design make complexity quieter.
Good defaults, calm feedback, and deliberate motion should reduce the amount a person has to decode.
Change the system when reality contradicts the story.
A neat narrative is not evidence. What the system actually did should be allowed to change the product and the belief behind it.
02 / Longer horizon
The world I keep moving toward
The world I keep moving toward is one where agents meet physical environments. Smart Manufacturing left me attentive to uncertainty, latency, energy, safety, and irreversibility—the things software can make easy to forget.
MIRAI-Setu deepened that interest through infrastructure, craft, and long time horizons across Japan. Physical AI is a direction I am studying, not a claim of current deployment or robotics expertise.
03 / Influences
What keeps shaping the work
origin
Learning the defaults by breaking them
Jailbreaking phones, rooting devices, and customizing PCs taught me to trace a system past its intended surface without mistaking tinkering for expertise.
community
Making technical depth inviting
HackByte made community-building concrete: invite people into serious technical work without lowering the standard.
Opens in a new tabplace
Japan, craft, and long horizons
MIRAI-Setu made infrastructure, safety, patience, and everyday attention to craft feel like engineering values rather than abstractions.
design
Design as quieter engineering
I care about what deserves motion, what should remain quiet, and how defaults can make powerful systems easier to trust.
04 / August 2026
Now
I am building Waldo, studying persistent agents and long-horizon evaluation, and looking to meet thoughtful builders and researchers working on how powerful systems can remain useful, legible, and on a person's side.
05 / Contact