Tools

What You'll Build

An on-demand engineer that lives in your phone and your laptop, so a one-person physical operation can do the kind of automation that used to require a payroll engineer and six-figure machinery.

Hiroki Tamiyasu farms broccoli in Hokkaido, the cold northern tip of Japan. He was a public servant before this. No farming background, no engineering background, none. He taught himself to farm, and he now oversees almost 250 acres of broccoli, pumpkins, green onions, and soybeans. At that scale the work is brutally physical, the operations are complicated, and good help is nearly impossible to hire out in the countryside.

The normal answer to that problem is to buy expensive proprietary equipment and pay specialized engineers to run it. He could not do that. So instead he treats an AI coding partner like an elite engineer who is always at his side. Concretely, he does four things with it. He photographs a sick broccoli plant in the field and gets a diagnosis in seconds. He built a bot that opens and closes his greenhouse vents remotely through a messaging app. He pulls satellite vegetation data and lays it over maps of his own plots to judge how each field is doing. And he used AI to understand GPS auto-steering well enough to realize he could build his own for a fraction of what a commercial rig costs. He also designs custom databases to run the farm.

You are not building a self-driving tractor this weekend. You are building the habit and the toolchain that let you, the owner of a physical operation, solve engineering problems yourself instead of waiting on a specialist you cannot find or afford.

Why This Works

For most of history, automation in a physical trade had a hard gatekeeper: you needed someone who could engineer it. That person was rare, expensive, and usually not interested in driving two hours to a farm or a job site. So the work stayed manual, or you bought a sealed proprietary box and paid whatever the vendor asked.

What changed is that the engineering knowledge is now on tap. You describe your problem in your own words, in your own language, and the AI translates it into a plan, the wiring, the code, the parts list. The farmer in this story does not know electronics. He does not need to. He knows broccoli, and he can describe what he wants. The AI fills the gap that used to require a salaried specialist.

It works especially well in hands-on operations because the hard part was never the physical world, it was the bridge to it. Sensors, relays, GPS modules, and small computers have been cheap and available for years. The missing piece was the person who could tie them together into something useful. That person now answers instantly, never quits, and costs a flat monthly fee instead of a salary.

And it flips the economics of every "you have to buy ours" pitch you have ever heard. A commercial GPS auto-steer system runs many thousands of dollars. Once you understand how it actually works, and the AI will teach you, you find out the underlying parts cost a small fraction of the sticker price. The markup was never the hardware. It was the engineering you could not do yourself. Now you can.

Prerequisites

Step-by-Step Setup

Step 1: Start With a Diagnosis, Not a Machine

Begin with the lowest-risk win, the one the farmer started with: photograph a problem and ask. Snap a picture of a diseased plant, a failing part, a strange reading, and feed it to the AI with a plain description. You get an answer in seconds and you learn how the tool thinks before you ever let it touch anything physical. This builds your trust and your skill at zero risk.

Step 2: Pick One Annoying, Reversible Job to Automate

Choose a single repetitive task where a mistake is cheap and fixable. Opening greenhouse vents is the farmer's example: if the bot gets it wrong, you walk over and fix it by hand. Avoid anything that moves heavy machinery or steers equipment for now. The goal of this first build is a working loop you control, not the most impressive thing possible.

Step 3: Have the AI Design the Wiring and the Code

Describe the job to your AI partner in full: what you want to control, what hardware you have or can buy, where it lives, how you want to trigger it. Let ChatGPT or your Claude-side tool lay out the parts list and the wiring, then let Codex or Claude Code write the actual software. Buy the cheap parts, wire them per the instructions, and load the code onto your small computer.

Step 4: Put the Controls in Your Pocket

Wire the trigger to a messaging-app bot so you can run it from your phone, the way the farmer controls his vents through a chat app. This is the part that makes it feel like magic and, more importantly, keeps you in the loop. You send a command, the thing happens, you can see it happened. No web of dashboards, just a chat you already check.

Step 5: Add Eyes Before You Add Hands

Bring in data that helps you decide before you build anything that acts on its own. The farmer pulls satellite vegetation data and overlays it on maps of his own fields to see which plots are struggling. Have the AI fetch a public data source, line it up with your operation, and present it plainly. Now you are making better calls with the same hands, which is most of the value with almost none of the risk.

Step 6: Only Then Tackle the Big, Risky Builds

The GPS auto-steering is the advanced tier, and you earn your way up to it. By now you understand your toolchain, you have shipped a few safe builds, and you trust your own judgment about what the AI gets right and wrong. Use it to understand the expensive proprietary system, then build your cheaper version, and prove it out slowly in a controlled space with a human ready to grab the wheel. Never go straight here from a standing start.

Adapting This for Your Business

The pattern is not about farming. It is about being the owner of a physical operation who never had an engineer and assumed they had to keep paying the specialist tax.

Gotchas and Tips

What This Replaces

Before this, automating a physical operation meant one of two things, and both were closed to a rural one-person outfit:

After this, an AI coding partner plays that engineer's role on demand for a flat monthly fee, and the cheap hardware that was always available finally becomes useful in your hands. A self-taught broccoli farmer with no engineering background now diagnoses crops from a photo, runs his greenhouse from his phone, reads satellite data over his own fields, and built his own GPS steering for a fraction of the commercial price. The larger point is the one worth sitting with: as AI spreads, even farmers in the most rural corners are gaining engineering skills that used to be completely out of reach. If he can, so can you.

Keep Reading