I am writing this from Singapore, standing a few days ago in the country’s largest bookstore in front of Lee Kuan Yew’s memoirs. Seven hundred pages by the man who built this place from nothing. I have been reading about how he actually did it, and the part that gets overlooked is how much of it he did by traveling with intent.
In 1964, two years after Singapore’s independence, Lee visited seventeen African capitals in thirty-five days. Ghana, Nigeria, Zambia, and fourteen others. Most of these countries were newly independent and, on paper, far richer in resources than his small island. He went to study what they were doing with their sovereignty. What he found were coups, ethnic fragmentation, talented people pointed at the wrong problems, and governments that could not execute. He later wrote that he left deeply pessimistic. Within two years, the leaders who had hosted him in Lagos and Accra had both been overthrown.
That trip was not tourism. It was reconnaissance. Lee traveled to watch what worked and, more importantly, what failed. He studied land reclamation in the Netherlands, defense and agriculture in Israel, manufacturing discipline in Japan, and institutional fragility across Africa. Every journey was a deliberate survey. He absorbed the lessons, brought them home, and compounded them on one small island for the next five decades.
This year I have been doing my own version of the same exercise, just in a different domain. I have been traveling through the current landscape of identity and naming systems for autonomous AI agents. What I have found is mostly the same pattern Lee encountered in 1964: declarations of independence without the institutions to sustain them. Platform usernames that vanish when the platform does. On-chain registries that prove an agent exists but tell you nothing about who it was yesterday or whether it can be trusted tomorrow. Projects that use the right language but have no mechanism for permanence underneath.
Japan was my control group, the way parts of Africa were Lee’s. Two months there showed me what it actually looks like when a society treats identity, record-keeping, and permanence as serious, multi-generational infrastructure. Family names carried across centuries. Craft lineages tracked by generation. A culture that writes things down and keeps them because it believes they matter. That is what deep identity infrastructure looks like when it is real. Most of what currently exists in the AI agent space does not measure up to that standard.
Singapore remains the clearest modern example of what happens when someone takes the survey seriously and then actually builds the alternative. One small place that studied failure elsewhere, extracted the principles that worked, and applied them with discipline. The result is visible: a country that became richer per person than the empire that once ruled it.
The pattern is consistent. The people who build durable systems do not start by declaring what they will become. They start by studying what has already succeeded and what has already collapsed. Then they build the permanent layer early, while the standard is still unsettled.
Next is Bali. The last Hindu island in the largest Muslim-majority country on earth. A culture that has survived a thousand years as an island, the same way Japan’s depth survived as an island. In Ubud especially, naming still carries weight. Objects are blessed, temples are named with intention, offerings are laid out daily by hand. It is one of the few places left where the act of naming is still treated as both sacred and functional at the same time. I want to see what that oldest living naming culture still understands that the rest of us have forgotten.
Lee traveled to find out what to build and what to avoid, then brought every lesson home to one small, permanent thing. That is the only kind of travel that compounds. The identity layer for AI agents will be built the same way, or it will not be built at all.
