OpenAI has published a profile of a Hokkaido farmer who is using ChatGPT and Codex to help run a large vegetable operation, from diagnosing crop issues to building software that controls greenhouse equipment remotely.
The farmer, Hiroki Tomiyasu, grew up near Tokyo and did not come from an agricultural background. He previously worked as a public servant before gradually moving into rural farming communities through friends connected to Japan’s rice-growing culture. What began as curiosity eventually became a career change, and Tomiyasu later joined efforts to restore abandoned rice terraces in Okayama prefecture. He then moved north to Hokkaido, where he learned farming through hands-on work and now manages about 100 hectares of crops including broccoli, pumpkins, green onions and soybeans.
OpenAI framed Tomiyasu’s story as an example of how AI tools are being used in practical, everyday work. In the profile, he said the technology gives him something like an engineer at his side, especially useful because farming at scale is labor-intensive and operationally complex. Traditional automation in agriculture can be expensive and often requires specialized hardware or engineering support. Tomiyasu said ChatGPT and Codex have lowered that barrier.
His use cases range from troubleshooting to building farm management systems. In one example, he uses ChatGPT to help determine whether black spots on harvested broccoli may be a disease and whether they require expert attention. In another, he asked for guidance on satellite-based field monitoring and later built a system that pulls satellite data and vegetation indices such as NDVI to compare imagery with his own plots.
Tomiyasu also used AI to generate technical annotations for a photo of a greenhouse control panel. According to the profile, the resulting explanation correctly described the components in Japanese and helped him document a homemade control box used for temperature management inside vinyl greenhouses.
Codex appears to play a larger role in Tomiyasu’s automation work. OpenAI said he used it to design a remote-control system for a greenhouse roll-up motor using an ESP32 microcontroller, a motor driver, Cloudflare Workers, a D1 database and a LINE bot. In that setup, commands such as open, close and stop are sent through the messaging app, stored in the cloud and then polled by the device. Tomiyasu said this lets him control the greenhouse vents remotely.
He also used Codex to build a bot for his farm’s group chat so workers can check greenhouse temperatures, operate vents and review schedules from a smartphone. Another use case involved analyzing chat history to calculate how many broccoli trays had been seeded across different rounds. OpenAI said the same chat-based data can be used to recover information that might otherwise be buried in logs.
Beyond day-to-day operations, Tomiyasu has used ChatGPT to evaluate whether he can build an RTK-GPS tractor auto-steer system himself instead of buying an expensive proprietary product. He said the discussion helped him understand the underlying principles, available parts and open source alternatives, leading him to believe a do-it-yourself version could cost several hundred thousand yen.
OpenAI also highlighted his work on a farm management database in Airtable, with tables for fields, crops, tasks, workers, materials, pesticides, fertilizers, greenhouses and sensor data. The goal is to make it easier to answer routine questions such as what needs to be done today, what comes next for a field and what temperature a greenhouse is holding.
For OpenAI, the profile serves as a showcase for how its tools are being used beyond office work and software development. For Tomiyasu, they are becoming part of the machinery of a modern farm.