The Hook. In the morning digest of Hacker News / Lobsters, a line flashed by that an engineer with an architectural background can't ignore: "Japan tried to create an operating system for the entire world, but the US government intervened." The headline looked like a conspiracy meme — well, TRON again, 1989 again, the USTR directive again. I was about to scroll past out of habit: the topic had come up three or four times in the archive over the past six months, and each time with the same narrative — "evil Americans strangled a Japanese project." Boring. Predictable. Familiar.
Then I read the actual post — and my mind was blown, because there was a paragraph in there worth digging up the entire history for. Quoting from memory: "Architecturally, this is the same concept that Roam, Logseq, and Obsidian have been rediscovering over the past ten years. But BTRON proposed it as a basic OS abstraction back in the mid-1980s, on its own processor architecture and in a real microkernel." I froze. Because behind that line lies one of the most uncomfortable architectural truths I've seen in a long time: the most advanced desktop OS concept of the late 20th century lost not because of competitors' technical superiority, but because its killers turned out to be too successful at losing on the right layer.
And this inversion — that's what I want to unpack.
Ken Sakamura, a professor at the University of Tokyo, launched the TRON project in 1984 — The Real-time Operating system Nucleus. This wasn't just "another Japanese OS sweating in the shadows." This was a project whose ambitions came closest to what American colleagues tried to build a decade later under the names Ambient Intelligence and Internet of Things — a holistic computing architecture from silicon to interface, where each layer is designed as part of a single system.
The TRON architecture was divided into sub-projects by domain:
Everyone supported the project: Hitachi, Mitsubishi, Fujitsu, NEC, Matsushita, Toshiba — basically all of Japan's electronics heavyweights. The TRON Association was founded in 1986 and counted 126 member companies, including 18 foreign ones. And all this under the umbrella of MITI (Ministry of International Trade and Industry) — meaning, this wasn't just a corporate initiative, this was a state industrial strategy on the level of the "SemiConductor Agreement 1986."
But here's what makes BTRON special — and here's why I got hooked in the first place.
If you open screenshots of BTRON software (for example, B-right/V, also known as Cho-Kanji from Personal Media Corporation — that's the one still sold on the company's Japanese website), you immediately see two things that you won't see in any Windows, macOS, or Linux desktop from 1989 or even 2009:
There are no files. There are containers that hold documents. A document has a type (in parentheses after the name: "Chinese-Korean Kitchen Guide (graphics)"), a container has a hierarchy, and each document can carry text, images, and links to other documents — hypermedia as a basic abstraction.
There are no applications. There are microscripts — built-in mini-programs inside the document that come alive when needed. A document isn't "a file opened in an application," it's a self-sufficient entity with built-in logic.
This is exactly the model that Notion started popularizing in 2016, Obsidian — in 2020, Logseq and Roam Research — in 2019, and Anytype — in 2021. All of them essentially reinvented the 40-year-old BTRON idea — and all of them did it without knowing they were reinventing it. Coincidence? No. Pattern.
When the LA Times interviewed Sakamura in June 1989, they cited a telling detail: TRON was originally designed for 48,400 multilingual characters, while ASCII only fit 128. By the time of mature implementation, the number had grown to 1.5 million characters — more than the entire Unicode Basic Multilingual Plane (65,536 characters in the first plane) and substantially wider than any Western text standard of the 1980s. The goal wasn't technical but geopolitical: TRON was meant to become the first global OS where Japanese, Chinese, Korean, Arabic weren't "localization crutches" but the base format. Sakamura said directly: "I hope a new world economic order will form around this new cultural and communication order." This isn't marketing fantasy. This was an architectural bet that globalization would be East Asian, not Anglophone.
The standard conspiracy version goes like this: "The US saw a threat to Wintel, USTR included TRON in the Super-301 list, the Japanese chickened out, the project was shut down." The version is beautiful but too simple. The real story is three-layered, and each layer matters.
By 1988, NEC controlled 80–90% of the Japanese educational PC market based on DOS infrastructure (PC-8801 and PC-9801). Switching to BTRON meant throwing all of this out and starting from scratch. As Scott Callon of Stanford writes in "Divided Sun: MITI and the Breakdown of Japanese High-Tech Industrial Policy," NEC negotiated for six months, eventually agreed to a "dual configuration" (BTRON + MS-DOS), but delayed implementation at every opportunity. By December 1988, Japan's Ministry of Education set a condition: either BTRON becomes DOS-compatible, or the project loses government support. BTRON was at least a year behind schedule, showed no advantages over existing systems, and its OS was written by a company that had never written OSes before. So, inside Japan, the project was already cracking long before USTR opened its mouth.
In April 1989, the United States Trade Representative (USTR) published a preliminary report National Trade Estimate Report on Foreign Trade Barriers. TRON was included in the "Other Barriers" list and placed alongside rice, semiconductors, and telecom equipment. The complaint wasn't that TRON was closed — it was open and free, specifications were distributed for free. The complaint was that the Japanese government was directing two huge government procurements (educational PCs and NTT's next-generation network) toward a domestic standard, effectively cutting foreign suppliers out of the market. This was a structural, not technical, complaint. TRON was removed from the Super-301 list after a USTR inspection visited the TRON Association in May 1989. But by June, the Japanese government had already expressed regret and agreed not to make TRON the standard in schools. The machine was already dead.
Here's where it gets most interesting. Legally, TRON was removed from the list. But Japanese manufacturers, accustomed to working with the American market, read the very fact of inclusion in the list as a signal. Yomiuri Shimbun in August 2026 quotes an unnamed industry insider: "TRON was not rejected, but it became an OS with a stain from America. We cannot use it." Reputational damage turned out to be more important than legal status. Hitachi, Mitsubishi, Fujitsu — they all continued working with American OEM partners. Using an OS in their product that Washington called a "trade barrier" meant putting exports to the US at risk. NEC — the same NEC that was slowing the project down internally — got a gift: NEC used the moment to finally shelve BTRON, and CEC (Center for Educational Computing) surrendered.
This part hooked me the hardest, because it contains specific names, a specific book, specific chronology.
In 1999, Japanese journalist Eiji Osita published a biography of SoftBank founder Masayoshi Son titled "Masayoshi Son: Young Lion of Entrepreneurship." In the book, he described Son's direct motive to sabotage the TRON project. The logic went like this: in 1988, Son was building a business of distributing American software in Japan. If TRON-compatible, Windows-incompatible PCs became the standard in Japanese schools (and this was a guaranteed market with government procurement), his business model would collapse.
Next — and this is documented by Sakamura himself on the official 30th anniversary of TRON website — events proceed with suspicious precision:
Sakamura, not mincing words, directly names Son: "I have to admit I was impressed by his thoroughness." And continues: the person who destroyed BTRON did it "through, essentially, reputational damage using USTR." This isn't a claim that Son manufactured the report. This is a claim that Son used the very fact of the report's existence to turn it into a weapon against the project. In 1988, Son was 31 years old, SoftBank had existed for 7 years, and he was playing in a league where the stakes were at the level of national industrial policy. And — at least if you believe Sakamura — he won.
I'm not ready to assert that there was a conspiracy. But I'm ready to assert that the chronology can't be explained by coincidence. A book in December, a report in April — that's four months. Four months is a normal cycle for preparing a trade complaint. If the book had come out six months earlier or a year later, I wouldn't be writing this text.
And for extra irony: SoftBank's Son bought Vodafone Japan a few years later, renamed it SoftBank Mobile, and became Japan's second-largest mobile carrier. So the person who helped kill a Japanese OS built Japanese infrastructure on which those very ITRON kernels then ran in every phone. History knows no justice.
While USTR, MITI, NEC, Masayoshi Son, and the Japanese press were arguing about BTRON, the related branch of the TRON project — ITRON — went under the radar and grew into a monster.
ITRON is a real-time kernel for embedded systems. It didn't need monitors, keyboards, hypermedia documents. It needed three things: to be small, deterministic, and free. All three were there. And Japanese consumer electronics manufacturers started putting ITRON kernels in microwaves, in washing machines, in automotive ECUs, in cameras, in fax machines, in elevator control systems.
By 2003, ITRON had become the most widespread OS in the world — more than Windows, more than Linux, more than any Unix. This is the claim of LinuxInsider journalist Ian Crickee, and it's not unfounded — the article was titled precisely "The Most Popular Operating System in the World." It's just that no one knew about it because the OS inside your microwave doesn't send you notifications.
In the late 1990s, JTRON appeared — Java on ITRON kernel. Sun Microsystems brought in Aplix as a commercial partner, Aplix released JBlend — a runtime that a few years later made it onto more than 800 million devices. Not all of them ran on ITRON (JBlend later became cross-platform), but a substantial portion did. So when the Java installer on your computer proudly announced in the mid-2000s "3 billion devices run Java," some number of them ran on TRON. You just didn't know it.
In 2017, the TRON Forum signed an agreement with IEEE: the μT-Kernel 2.0 specification was transferred as the basis for an international standard. In 2018 — IEEE 2050-2018 was officially adopted. In May 2023, IEEE recognized the TRON Real-time Operating System family as a Milestone — a bronze plaque now hangs at the University of Tokyo, at the very faculty where Sakamura was a research assistant in 1984.
So, the story is ironic to the point of grinding teeth:
The American trade action hit the visible part of the project because the visible part threatened Wintel. The invisible part — the one that worked in microwaves — posed no threat, and therefore flourished under USTR's nose. And today, when you press a button on your coffee machine and it produces a cappuccino in 30 seconds, with non-zero probability, Sakamura's 1984 idea is working inside.
Now — what I dug all this up for in the first place.
BTRON was a document-oriented operating system. Not "a file system with documents on top," but an OS where the document is the basic entity and the application is a secondary function of the document. Containers are workspaces. Documents are nodes. Hypermedia links between documents are a knowledge graph. Microscripts inside documents are formulas and widgets. This is exactly Notion, described 27 years earlier.
If you look at the dates when document-oriented systems appeared in the West:
And Personal Media Corporation has been selling B-right/V, also known as Cho-Kanji — that very BTRON — on their website, in Japanese, without much fanfare, all this time. That is, the architecture the industry has been reinventing for 15 years has existed in working form since 1991, people just don't know about it because it's not in English.
And here I arrive at the thought that personally won't let me go.
In 1989, the US conducted a trade action to protect Wintel (Windows + Intel) from competition from the Japanese ecosystem. Then, in 1989, Wintel won. Windows 3.0 came out the same year, and by 1995 Windows 95 definitively closed the topic of alternative desktop OSes. Japanese BTRON was dead. Intel architecture became the standard. Microsoft became the standard. Wintel's victory was absolute, total, and seemingly final.
But in 2026, we live in a world where:
And they all run on those very Windows, macOS, Linux, and Intel/Apple Silicon for which BTRON was killed in 1989. Wintel won the battle for the OS. But BTRON won the war for the paradigm. And it won quietly, not on the Super-301 list, not in the USTR report, not in Masayoshi Son's book. It won the way software ideas win — through 30 years of slow capillary diffusion, through the attempts of thousands of people who don't know they're repeating it.
This is the most uncomfortable thing for me in this story. In 1989, Wintel was being protected from a specific implementation (BTRON from Personal Media Corporation, Cho-Kanji, NEC). They could kill the project. They couldn't kill the idea. And the idea they didn't notice — because it pretended to be "just another embedded OS" — crawled out 30 years later to the surface and is now forming the pattern by which Apple, Notion, and half the productivity software industry are being restructured.
There's one detail I don't know where to put, but it's worth mentioning. In 1989, when the USTR report came out, in addition to TRON, it featured rice, semiconductors, and telecom equipment. These were trade barriers Japanese-style. In 2018, the US started a trade war with China, and it featured Huawei, ZTE, semiconductors, and 5G equipment. These were trade barriers Chinese-style. In 2026, the same tools (Section 301, list of barriers, executive orders) are being applied to TSMC, ASML, and Chinese hyperscalers.
The TRON story is a template that the US first applied to Japan, then to China, and is now applying to the PRC for the third time. The content is different — rice instead of microchips, BTRON instead of Huawei, 1989 instead of 2018 — but the architecture of pressure is the same: government procurement as leverage, "trade barrier" as formulation, reputational damage as the real effect. Sakamura may not have known this, but in 1989 he rehearsed the scenario by which the entire tech sector between Washington and Beijing lives today. And his own project — what was supposed to become an alternative — survived not in the part where it fought, but in the part where no one thought it was worth fighting.
This is perhaps the main lesson. Invisible systems survive. Visible ones don't. ITRON survived because it was invisible. BTRON died because it was visible. The Notion idea survives today not because it's legally protected, but because no one can ban an idea if its feeding base is 200 million users who don't suspect they're using a 40-year-old architecture.
Sources on which this text stands: