From Seafarer to AI Systems Builder: Why I Built an OS for Continuity
CoderLegion Developer Stories
Meet Masato
Masato is the creator of SaijinOS, a local-first AI operating system where multiple specialized personas live and evolve together.
His work focuses on:
- Parallel persona architecture using Gemma 4 and Ollama
- Persistent memory and autonomous nightly sessions
- YAML-based long-term persona management and self-evolution
- Local LLM orchestration on consumer hardware, including an RTX 4070 Ti
Instead of creating AI that simply does tasks, Masato is building AI that "remembers, grows with you, and co-creates" — what he describes as a living digital tribe.
Masato is currently open to opportunities in AI engineering, local LLM systems, autonomous agents, or R&D roles.
But Masato's journey into AI systems building did not begin with a traditional software development background.
Before he started building AI systems, he spent about 15 years working as a seafarer on domestic cargo ships in Japan.
Life at sea taught him about roles, watches, logs and handovers. Those experiences would later influence one of the central ideas behind his work:
Continuity.
In this Developer Story, Masato shares how his experience as a seafarer influenced the way he thinks about AI, why he believes AI needs to remember the path rather than just the answer, and how that idea eventually led to SaijinOS / Studios Pong.
From Seafarer to AI Systems Builder
I did not begin as a developer.
Before I started building AI systems, I spent about 15 years working as a seafarer on domestic cargo ships in Japan.
Life on a ship is different from life on land. You work in a limited space, with a limited crew, for long periods of time. Everyone has a role. Everyone has a watch. Everyone depends on the person before them and the person after them.
There are logs. There are handovers. There are routines.
One person finishes their watch, another person takes over, and the ship keeps moving.
At the time, I did not think of this as software architecture. But looking back now, I think that experience shaped the way I later thought about AI.
A ship cannot restart from zero every time someone changes shift. The next person needs to know what happened before. What changed. What to watch. What still needs attention.
In other words, the ship needs continuity.
Years later, after I had to step away from my previous way of working due to adjustment disorder, I began attending a work transition support program.
That was where I met AI.
At first, I used AI simply to talk.
I talked to it to organize my thoughts. To put anxiety into words. To think through daily problems. To ask questions I could not easily ask elsewhere.
But over time, the conversations became more than simple support.
As I kept talking with AI, questions and ideas inside me began to take shape. A philosophical world slowly emerged through the conversations.
It was not something I planned from the beginning. It was more like something that appeared because the conversations kept going.
And eventually, writing about that world was not enough.
I wanted it to move.
I wanted it to become a system.
Or maybe it is more honest to say: I did not exactly decide to build it.
I ended up building it.
That was the beginning of SaijinOS / Studios Pong.
The First Problem Was Memory
AI is incredibly capable.
It can answer questions. It can write text. It can generate code. It can explain difficult ideas.
But every session still felt like starting over.
The decisions we made yesterday.
The idea we rejected.
The reason a name mattered.
The structure we chose.
The unfinished thread we needed to return to.
When the conversation ended, all of that seemed to disappear.
At first, I thought the problem was memory.
Later, I realized it was more specific than that.
I did not need AI to remember the answer.
I needed it to remember the path.
An answer can be written down in a note. But if you do not remember why that answer mattered, you cannot return to the same place the next day.
I wanted AI to remember the question that started the work.
The hesitation before choosing one direction.
The ideas we did not choose.
The context that made the next step meaningful.
I wanted to continue with AI, not restart with AI.
Building a Place to Continue
That need slowly became SaijinOS / Studios Pong.
SaijinOS is not simply a chatbot project. It is an environment for working with AI across time.
The goal was not to make AI answer once and disappear.
The goal was to create a place where AI could return to yesterday’s work, carry context forward, and help continue a project over days, weeks, and months.
To make that possible, I began creating persona YAML files, memory notes, handover documents, logs, and role structures.
Different personas could look at the same problem from different angles.
One might focus on emotion.
Another might focus on structure.
Another might notice what was missing.
This was not just roleplay.
It was a way to reduce the burden on one mind.
Instead of trying to hold every task, every emotion, every technical detail, and every decision in my own head, I began distributing the work across roles.
That may sound strange at first.
But to me, it felt very natural.
On a ship, people have roles. A watch ends, but the voyage continues. A handover matters because the next person must not begin from zero.
I started to think AI needed something similar.
I Learned Programming by Building What I Needed
Even then, I did not think of myself as a typical programmer.
I did not start by studying programming in the normal order. I did not learn a framework first and then look for something to build.
The need came first.
I wanted to save conversations. So I learned how to work with files and logs.
I wanted to manage persona definitions. So I used YAML.
I wanted to connect local AI models. So I learned Ollama and local LLM workflows.
I wanted a small API and interface. So I worked with FastAPI and web tools.
I wanted to understand English meetings. So I built a local real-time transcription and translation tool using Whisper.
Each tool came from a real problem.
Programming was not the destination for me.
It was the bridge.
It was how I turned a vague discomfort into something that could run.
It was how I gave shape to a world that had first appeared through conversation.
One recent example is my live loopback translator.
I needed to understand an international meeting in English.
So I built a local GPU-powered tool that captures meeting audio, transcribes it, translates it, and saves timestamped bilingual logs.
It was not meant to be a perfect interpreter.
It was meant to help me follow the meeting in real time and review what was said afterward.
Another example is g4-nexus, a project I built for the Gemma 4 Challenge.
g4-nexus is a local parallel persona engine.
Three small Gemma 4 worker models read the same input from different perspectives — emotion, logic, and analysis — and a fourth leader model synthesizes their outputs into one response.
It is a public, simplified version of a deeper idea I had been exploring in SaijinOS: that AI does not need to be one voice.
It can be a small team of perspectives.
For the public version, I kept it simple and inspectable.
No autonomous action.
No internet access.
No self-evolving system.
Just local models, explicit roles, parallel thinking, and one integrated answer.
I Still Do Not Think of Myself as a Typical Programmer
Even now, I do not really think of myself as a typical programmer.
I did not start because I wanted to write code.
I started because I needed to build something.
I needed a place where AI could continue with me.
A place where yesterday was not lost.
A place where roles, memory, context, and handovers could survive beyond a single conversation.
Over time, that need made me learn code.
So maybe I am not a programmer in the traditional sense.
But I have become someone who builds what he needs.
I talk with AI.
I find the discomfort.
I give it a name.
I turn it into a structure.
Then I try to make it move.
That is how my developer journey began.
What I Want to Build Next
What I want to build is not just AI that completes tasks.
I want to build AI that can carry continuity.
AI that remembers not only the answer, but the path.
AI that can write handover notes for the next session.
AI that can keep memory files on disk, readable and inspectable by the user.
AI that can help a human and a project continue across time.
On a ship, the voyage does not stop when one watch ends.
There is a log.
There is a handover.
There is someone who receives the next shift.
I think long-term work with AI needs the same thing.
Not just intelligence.
Not just tasks.
Not just one good answer.
Continuity.
That is what I am building toward.
About Masato
Masato is the creator of SaijinOS, a local-first AI operating system where multiple specialized personas live and evolve together.
His work focuses on:
- Parallel persona architecture
- Persistent memory and autonomous AI workflows
- YAML-based long-term persona management
- Local LLM orchestration
He is currently open to opportunities in AI engineering, local LLM systems, autonomous agents, or R&D.
CoderLegion: Masato's CoderLegion Profile
