In 1965, Nintendo was not the company we know today. It was a small Japanese firm mostly producing hanafuda playing cards. That year, they hired a young engineer named Gunpei Yokoi.
Yokoi was 24. He had graduated from Doshisha University with a degree in electronics. By every account, including his own, he had not been a distinguished student. Nintendo president Hiroshi Yamauchi assigned him to the only role available: maintaining the assembly-line machines that produced the playing cards.
It was the most unglamorous engineering position in the building. Yokoi’s career, by every conventional measure, was over before it had begun.
What happened over the next thirty years is one of the most important counter-intuitive case studies in the history of innovation — and the philosophy he articulated is the single most important framework available to senior leaders navigating AI transformation in 2026.
The factory floor as vantage point
Maintaining the playing card machines gave Yokoi an unusual vantage point. He was surrounded, every day, by hardware that he himself had not designed. Belts. Pulleys. Levers. Off-the-shelf motors. Photoelectric sensors that had been available for a decade.
What Yokoi began to notice — and what specialists with their eyes on the frontier were missing — was that there was an enormous landscape of mature, inexpensive, deeply-understood technology sitting all around him. Reliable. Boring. Overlooked by everyone whose career depended on chasing the next breakthrough.
The mature technology had three properties that mattered. It was cheap. It was debugged. It was available. Frontier components had none of these properties.
Yokoi realized that this created an enormous, uncontested opportunity space. He could not compete on the frontier. But he could combine mature, well-understood parts in novel ways to solve problems that frontier-chasing engineers weren’t even thinking about.
The Ultra Hand
In 1966, Yokoi built a simple mechanical toy out of parts he had access to in the factory. An extending claw, made from interlocking parallelogram mechanisms. The mechanism had been understood by engineers for a century. Yamauchi saw the prototype and asked him to commercialize it.
The Ultra Hand launched in 1966. It sold approximately 1.2 million units in Japan in its first year. Yokoi articulated the underlying philosophy with a phrase that has been taught to Japanese engineering students ever since:
枯れた技術の水平思考 — Kareta gijutsu no suihei shikō. “Lateral thinking with withered technology.”
The word kareta — translated as “withered” but better understood as “matured” or “seasoned” — refers to technology that has been around long enough to be cheap, reliable, debugged, and ignored.
The Game Boy
The deepest test of Yokoi’s philosophy came in 1989, with the Game Boy.
The Game Boy launched in April 1989 with a Sharp LR35902 processor — a Z80 derivative whose architecture was first introduced in 1976, thirteen years old at launch. A 4-shade monochrome green LCD, compared by reviewers to rotting alfalfa. In every measurable specification, this was technology that should have been obsolete years before launch.
Two direct competitors launched at almost the same time. The Atari Lynx (September 1989) featured a full-color backlit LCD and a 16-bit processor. The Sega Game Gear (October 1990) had a full-color LCD and dedicated sound hardware. Both were objectively superior to the Game Boy in every technical specification.
The Game Boy outsold both combined by roughly ten to one. Approximately 118 million units sold worldwide.
What Yokoi understood that the competitors didn’t
Battery life. The Game Boy’s monochrome LCD and decade-old processor consumed almost no power. Four AA batteries produced 30+ hours of gameplay. The Lynx got 4 to 6 hours on six AA batteries.
Price. The Game Boy launched at $89.99. The Lynx was $179. The Game Gear $149.
Durability. The Game Boy could survive being dropped. The color competitors were fragile.
The game library. The Game Boy’s mature hardware was easy for developers to design for. Tetris. Pokémon Red & Blue. The Legend of Zelda: Link’s Awakening. Super Mario Land.
Every one of these advantages came from Yokoi’s commitment to mature, well-understood technology. The Lynx and Game Gear engineers built impressive machines that lost the market by failing to deliver what humans actually cared about.
The 2026 application
The reason I am writing about Yokoi in 2026 is that almost every senior leader I work with is making the Lynx mistake about AI.
The implicit assumption underneath most enterprise AI strategy in 2026 is that progress requires using the most advanced AI technology available. The newest models. The latest frameworks. The most cutting-edge architectures.
This assumption is correct for a narrow subset of AI use cases. It is wrong for the majority of high-value AI use cases inside an enterprise.
The customer service workflow handling 80% of routine queries. The Yokoi approach uses a smaller, mature, much cheaper model that handles the routine cases reliably at a fraction of the cost.
The internal AI assistant for knowledge workers. The Yokoi approach uses a stable, well-understood model with focused integration into specific workflows, deployed with clear guardrails and predictable behavior.
The document summarization pipeline. The Yokoi approach uses a mature retrieval-augmented generation pattern deployed on a smaller, cheaper model.
The regulated business AI workflows. The Yokoi approach uses stable, well-understood models with established compliance characteristics.
In every one of these cases, the Yokoi move beats the frontier move.
What this means for executive practice
One: most of your AI portfolio should run on withered technology. The frontier should be reserved for the small minority of use cases where the marginal capability gain genuinely matters.
Two: hire for withered technology expertise, not just frontier expertise. The engineers who produce the most enterprise AI value are often not the ones chasing the frontier. They are the ones who deeply understand mature, deployable models.
Three: focus on what actual humans care about, not on technical specifications. Battery life. Durability. Price. Reliability. Consistency. None of these are technical specifications. All of them are attributes of human experience.
The closing thought
Gunpei Yokoi died on October 4, 1997, at age 56, in a traffic accident. He left behind a philosophy: 枯れた技術の水平思考.
The philosophy says that progress is not the same as the frontier. The frontier is where specialists fight each other for marginal performance gains. Progress often comes from someone with the discipline to use the mature, boring, overlooked technology in a new way.
The Game Boy beat the Lynx by ten to one. Your AI portfolio, if you build it with Yokoi’s discipline, will beat your competitors’ frontier-chasing portfolios by similar margins. Not because your technology is better. Because you knew what the humans actually cared about.
The world has changed. The leaders who notice will be the ones the next decade is built around.
