KASANEGI

01 · Research

Searching all of Japan for undiscovered meteorite craters

Building on the latest red relief image map of all of Japan, this research uses machine learning to train AI on the world's known impact structures, so that the AI can identify undiscovered meteorite craters across the terrain of the entire country. It works at a speed, and in a way, that checking by eye and overlaying data by hand cannot match.

In Japan, Oikeyama in Nagano Prefecture has been reported as a meteorite crater. The country's steep, forest-covered mountains are thought to hide craters that remain undiscovered. Identifying craters in imagery covering all of Japan requires advanced AI built on machine learning.

AI agents carried out preparation on a scale no human team could reach. The AI learns from the world's impact structures. We have compiled 354 of them, across 67 countries and regions, into a single inventory, and linked 2,011 papers to 240 of those structures. For 48 known craters, we have made red relief image maps from publicly available 1 m elevation data. We have also identified the calculation behind Japan's nationwide red relief image map down to its formulas and constants, and reproduced it at Oikeyama to within one colour level. This lets us compare terrain worldwide and in Japan by the same measure.

Beyond terrain, we assess the quality of Japan's nationwide gravity and airborne magnetic surveys against primary sources and the actual data. Candidates the AI puts forward are verified by our founder, a certified mineral appraiser and geology specialist, who examines the rocks in the field. The same method extends to terrain worldwide, to the surface of the Moon, and to the discovery of kofun burial mounds.

354

impact structures worldwide in our inventory, across 67 countries and regions, including 226 confirmed sites

2,011

papers linked to 240 structures in the inventory

48

known craters made into red relief image maps from 1 m terrain data

03 · Data API

Japan's public data, in a form AI agents can use directly

We turn tables and registers published by Japan's public institutions into APIs that AI agents and software can use in a single call. Sources include the National Tax Agency, the Japan Health Insurance Association (Kyokai Kenpo), the Ministry of Health, Labour and Welfare, the Cabinet Office, and gBizINFO from the Ministry of Economy, Trade and Industry. Since before our founding, Kasanegi has developed a family of these data APIs and published them on kasanegi.com, on a marketplace for developers and over MCP.

We are now integrating that family of data APIs into the foundation that brings all of Kasanegi together. In the meantime, so that you can work with the real thing, the ten APIs of JP Local Pack are open free of charge, with no sign-up. JP Local Pack is one approach within the family: public data for back-office work in Japan.

Even this one example is exact: the withholding income tax API holds all 1,856 cells of the National Tax Agency's monthly withholding table and matches the official table cell for cell.

Beyond the APIs we have opened, we hold further data APIs: matching and enriching company records, corporate disclosures such as annual securities reports, and searching grants and judging eligibility, all built to connect public data directly to an AI agent's decisions. As our AI resources grow, we open more of them.

Our data APIs are built so that AI agents can find and use them on their own. With an MCP server, OpenAPI and llms.txt in place, an AI agent receives the usage instructions and endpoints, and can put the APIs to work directly.

1,856

cells of the National Tax Agency's withholding tax table, every one matched, even in JP Local Pack, one example from the family

5 million

entries in the qualified invoice issuer registry, which an AI agent can look up in one call

10

APIs open free of charge now, as one example from the family, with MCP support