Language, Curiosity, and Life
April 12, 2026 — Masato Hagiwara
Over the past few months, my cancer has progressed significantly, and the latest treatment did not work. After discussions with my doctors and my family, I have decided to move to hospice care. Hospice care is end-of-life care that focuses on comfort and quality of life rather than continuing aggressive treatments. I feel it is important to share my “last words” while I still have the time and clarity to do so, rather than waiting until the final days, when I may not be able to express what truly matters. I am grateful to my current employer, Earth Species Project (ESP), for allowing me to continue working full-time even while in hospice care.
I have been living with cancer for about three years. I was diagnosed with lung cancer in March 2023. Since then, I have kept a private daily diary. Much of it is practical—a record of medications, doses, and symptoms—but it also captures major moments in my treatment, including emergencies, like the time I was taken by 911 due to severe bleeding from a stomach ulcer. Over time, I found that even writing just one line a day is enough to create a meaningful record. It helps you track your condition, but more importantly, it helps you make sense of what you are going through.
Language
Language has been a central thread in my life. In many ways, it is where everything started.
After finishing my PhD, I chose my first job at Baidu partly because it involved working across languages. It simply sounded fun to go to China and work in an environment where I could use two or more languages every day. That decision shaped my life more than I expected—it was also where I met my wife. Living and working in China exposed me to a very different way of thinking and working compared to Japan, which was an eye-opening experience.
My wife is naturally multilingual. She grew up speaking Hakka, a Chinese language spoken in southern China, along with Cantonese, which is widely used in nearby cities, and Mandarin, the standard language taught in school. Later, she learned English and Japanese in academic settings. What struck me most was how languages can change dramatically even within a short distance—around her hometown, you can drive for an hour and encounter entirely different linguistic environments. This is not necessarily true everywhere in mainland China, but it reflects the deep linguistic diversity that exists in many regions. China is, in that sense, a truly linguistically diverse country.
My own Chinese language skills improved largely because of her. Early on, she deliberately avoided acting as a translator between me and her family. Even though she could easily do it, she “forced” me to communicate directly with her parents. It was uncomfortable at first, but it turned out to be the most effective way to learn.
When we had our first child, one of the earliest decisions we made was to adopt the OPOL (one-parent, one-language) approach. I speak Japanese to our children, while my wife speaks Mandarin. Because I understand Mandarin and she understands Japanese, we can naturally join each other’s conversations. The children use English with one another, as it has become their primary language for school and social interactions. This approach felt natural to us, and we have continued it for over ten years. So far, it has been working well. We also built small habits around language. For example, we watch a movie together every week, rotating the language—Japanese, Chinese, and English. It not only reinforces language exposure but also broadens the range of movies we watch.
One of the reasons we moved from Tokyo to New York City was that a professor I knew was starting a new research lab at the Rakuten Institute of Technology. The work involved natural language processing tasks like named entity recognition and machine translation—again, connected to language.
After Rakuten, I joined Duolingo, which felt like an obvious step for someone interested in language learning. After that, I worked as a freelancer for a while before receiving an unexpected email from one of the co-founders of Earth Species Project (ESP). The project he described—using AI to decode animal communication—was one of the most interesting ideas I had ever heard. That is how I joined ESP, first as a contractor and later as a full-time researcher.
Looking ahead, I sometimes feel conflicted about the future of language learning. For casual situations like travel, it is becoming less necessary to learn a language deeply. Translation apps and AI tools already allow people to communicate with minimal effort. I have seen many real-world situations where people rely entirely on their phones to bridge language gaps—for example, store clerks and foreign travelers interacting smoothly without a shared language, simply by using translation apps.
At the same time, for professional work—whether in business or academia—serious language learning remains essential. If you want to operate at a high level, you still need deep proficiency in languages like English or Mandarin. I think of this as a kind of “binarization” of language learning: casual use cases are shrinking, while high-level use cases remain strong.
Curiosity
Curiosity is probably the most important driving force in my life, and much of it comes from my father. He had a simple way of thinking: “If it’s something other people can understand, why can’t I understand it?” That mindset stayed with me. It made me believe that most things are, in principle, learnable. Many of my interests—music, programming, design—started from that belief. Looking back, a large part of my curiosity comes from him.
Curiosity also showed up early in more tangible ways. I started playing music when my father brought home an organ for free when I was a kindergartener. I kept practicing, and by middle and high school I was playing in bands across different genres, including brass and rock. Music became not just a skill, but a way to connect with others—classmates, friends, and collaborators.
Later, I joined Duolingo when it had around 40 employees, and I left when it had grown to over 100. I also joined ESP when it was still a very small organization, and saw it grow into a much larger team. Over time, I realized I am drawn to places in their early stages, when ideas are still forming and the direction is not yet fixed.
After leaving Duolingo, I worked as a freelance AI engineer for about a year and a half. I wrote about that experience in an essay called My First Year as a Freelance AI Engineer, which unexpectedly reached the front page of Hacker News and attracted a fair amount of attention. More importantly, it was a very positive chapter in my life. I have often thought of myself as being closer to a creative and curious artist than to a conventional engineer or researcher, and freelancing felt like a natural expression of that side of me.
Life
There are certain pieces of advice that have stayed with me for many years.
My parents always told me to go to sleep by 9 p.m. Looking back, I think that was good advice—for both health and academic performance. One of the last conversations I had with my father, while he was dying in the hospital, was about music. He told me that I should listen to as wide a variety of music as possible. I still think that was one of the best pieces of advice he ever gave me.
At the same time, I learned that not all advice should be followed equally. My parents’ generation could not have imagined many of the careers that exist today. I doubt many parents twenty years ago were telling their children to become AI engineers. That is one reason I do not think young people should follow their parents’ career advice too literally. Parents can pass on values, habits, and ways of seeing the world, but they cannot fully predict the future.
During my freelancing time, I also started the blog State of AI Guide in August 2020, around the time GPT-3 appeared. It was a Japanese, subscription-based blog that covered recent developments in AI—summaries of interesting papers, conference proceedings, and new ideas I wanted to understand better. It began as a collection of my own notes, but over time it grew into something much larger, eventually generating enough income to correspond to a full-time job. I stopped writing it soon after I was diagnosed with lung cancer in March 2023, just before AI exploded into mainstream attention with ChatGPT and everything that followed.
That experience taught me one of the most important lessons of my illness: do not make major life decisions too quickly after a diagnosis. When people hear the word “cancer,” it is natural to panic and rethink everything at once. But the mind is not at its best in that state. For anyone who may face big life changes in the future, one practical piece of advice I would give is this: unless there is an immediate reason, do not rush to quit your job or drastically rearrange your life right away. Give yourself time to understand the reality before making irreversible decisions.
One pattern I’ve noticed in my life is that I tend to follow people rather than predefined paths. When I joined Earth Species Project, I was deeply impressed by the people who were already there—their thinking, their work, and their ambition. Over time, I’ve found a simple and reliable way to decide whether a workplace is a good fit for you: just talk to the people who work there. Especially in technical fields, it becomes clear very quickly whether someone truly understands what they are doing. Moreover, I focus on concrete signals rather than abstract narratives. Not just funding or reputation, but metrics that reflect real engagement—daily active users, GitHub stars, or anything that shows people genuinely want what is being built. In simple questions: are they making something people truly want? That is what makes an organization take off.
Another idea that has influenced me is Derek Sivers’s essay: No “yes.” Either “HELL YEAH!” or “no.” As someone who is easily distracted and constantly full of ideas, I have found this principle extremely useful. It has helped me separate the things that are merely interesting from the things that truly matter to me.
One example is Recto, a project I published in August 2025, when my health was relatively stable. Under normal circumstances, I might have treated it as just one of many side projects. But because time had become more visibly limited, it felt more like a bucket-list project—something I deeply wanted to bring into the world while I still could. I would recommend that kind of thinking to others as well. We all have things we say we want to do “someday,” but sometimes the right question is whether that someday should actually be now.
Meanwhile, I have also learned the downside of having too many ideas. I have always had more projects in mind than I could realistically complete. As someone who wants to create things, that can feel energizing, but it can also dilute attention. Too many half-finished projects can pull energy away from what matters most. I am still learning this lesson, and I do not claim to have mastered it.
My illness made these questions much more concrete. In July 2025, I was diagnosed with leptomeningeal disease (LMD) and went through proton therapy, along with side effects such as seizures and headaches. It was one of the hardest periods of my life. But it also led to another major lesson: do not believe prognosis too easily.
When I was diagnosed with LMD, I did what many people do: I searched online. What I found was grim. The message seemed to be that people in my situation often had only weeks or months to live. But reality turned out to be more complicated than that. My treatment with a targeted therapy (Tagrisso) was still working, and I lived much longer than those expectations suggested. That experience changed the way I think about medical predictions. Statistics are real, and prognosis matters, but an individual life is not reducible to an average. I wrote more about this in another essay, because it became one of the most important lessons of my cancer journey.
I would not have been able to come this far without the people around me. My family has carried me in ways that are hard to fully describe—through ordinary days, difficult treatments, and moments when things felt uncertain. My wife, especially, has been at the center of everything, holding together both the practical and emotional parts of our life with a strength I deeply respect. My children have given me something even more important: a reason to keep going, and a constant reminder of what really matters.
I am also grateful to my close friends, who have supported me in both visible and invisible ways—through conversations, messages, and simply being there. I also want to take a moment to express my deep gratitude to all who have supported me throughout this journey. It is difficult to fully describe what that support has meant to me. In many moments, it was not something visible or dramatic, but something quiet and constant: being there, listening, staying connected. That has made all the difference.
At the same time, I do not think of this as a “goodbye.” If anything, this is a continuation—just in a different form. Through these words, through the recordings, through the ideas and memories I leave behind, I hope the conversation continues.
This article is published under Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License