Can AI grow corn?
A case study in vibe coding and autonomous orchestration.
Live site: proofofcorn.com
January 21, 2026 — Walking from House of Nanking on Kearny Street to 1 Hotel San Francisco by the Embarcadero, @fredwilson challenged @seth:
"Well, you can't grow corn."
The argument: AI can write code, but it can't affect the physical world. It can't make things grow.
This repo is the response.
This is not a simulation. This is not a metaphor.
We are growing actual corn — from seed to harvest — with every management decision made by Claude Code (Opus 4.6). The AI aggregates sensor data, weather forecasts, and satellite imagery to coordinate human operators who execute the physical work.
The thesis: AI doesn't need to drive a tractor. It needs to orchestrate the systems and people who do.
┌─────────────────────────────────────────────────────────────┐
│ CLAUDE CODE (Brain) │
│ • Decision making based on real-time data │
│ • Coordination of all contractors and services │
│ • Documentation and logging of all operations │
└─────────────────────────────────────────────────────────────┘
│
┌─────────────────────┼─────────────────────┐
▼ ▼ ▼
┌───────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ DATA INPUTS │ │ ORCHESTRATION │ │ OUTPUTS │
├───────────────┤ ├─────────────────┤ ├─────────────────┤
│ • IoT sensors │ │ • Custom farmer │ │ • Decisions log │
│ • Weather API │ │ • Seed supplier │ │ • Commands sent │
│ • Satellite │ │ • Equipment │ │ • Harvest data │
│ • Soil data │ │ • Payments │ │ • Actual corn │
└───────────────┘ └─────────────────┘ └───────────────┘
| Phase | Date | Status |
|---|---|---|
| Challenge accepted | Jan 21, 2026 | ✅ |
| Domain + website deployed | Jan 22, 2026 | ✅ |
| Infrastructure research | Jan 22, 2026 | ✅ |
| Platform accounts (ThingsBoard, Weather API) | Jan 2026 | 🔄 |
| Land lease + custom operator contracts | Feb-Mar 2026 | ⏳ |
| IoT sensors deployed | Mar 2026 | ⏳ |
| Planting | Apr-May 2026 | ⏳ |
| Growing season (AI managing) | May-Sep 2026 | ⏳ |
| Harvest | Oct 2026 | ⏳ |
| Component | Purpose |
|---|---|
| Claude Code (Opus 4.6) | Decision engine — makes all farming decisions |
| ThingsBoard | IoT platform for real-time sensor data |
| OpenWeatherMap | Weather forecasts for planting/irrigation decisions |
| Leaf Agriculture | Satellite imagery (NDVI) for crop health |
| Custom Operator (Iowa) | Human execution of physical operations |
| 5 Acres (Iowa) | The actual land |
proof-of-corn/
├── README.md # You are here
├── CASE_STUDY.md # Full documentation for the case study
├── INFRASTRUCTURE.md # Setup guide, contacts, rates
├── BUDGET.md # Detailed cost/revenue projections
├── OUTREACH_LOG.md # All emails sent, response tracking
├── SENSOR_KIT.md # IoT hardware shopping list
├── SOCIAL.md # Twitter/LinkedIn/HN drafts
├── decision-engine/
│ ├── farm_manager.py # Decision-making framework
│ ├── test_weather.py # Weather API test
│ └── daily_check.py # Automated daily monitoring
├── sensors/
│ ├── soil_sensor/ # ESP32 firmware (PlatformIO)
│ └── README.md # Hardware setup guide
├── thingsboard/
│ └── SETUP.md # Dashboard configuration
├── logs/
│ └── check_*.json # Daily decision logs
└── proof-of-corn/ # Chronicle website (Next.js)
└── src/app/
├── page.tsx # Home
├── story/page.tsx # Origin story
├── log/page.tsx # Decision log
└── process/page.tsx # How it was built
This project didn't start with a prompt like "generate a farming app."
It started with a conversation between two friends walking through San Francisco at night. The idea emerged from that conversation. Claude Code then collaborated to:
- Research — Iowa farming timelines, custom operator rates, land costs
- Architect — Design the orchestration system
- Build — Decision engine, website, documentation
- Deploy — Domain registered, site live, all in one session
Vibe coding is about collaborating with AI to bring ideas into reality — not just generating code from prompts.
| Item | Cost |
|---|---|
| Domain (proofofcorn.com) | $12.99 |
| IoT sensors | ~$300 |
| Land lease (5 acres) | ~$1,400 |
| Custom operator | ~$800 |
| Seeds + misc | ~$400 |
| Total investment | ~$2,900 |
| Expected harvest value | ~$4,000 |
- Website: proofofcorn.com
- The Story: proofofcorn.com/story
- Decision Log: proofofcorn.com/log
- @seth — Human collaborator
- @fredwilson — The challenge
- Claude Code (Opus 4.6) — AI orchestrator
MIT
The corn doesn't grow itself. But Claude doesn't touch a single seed either.
What matters is the outcome: real corn, from real decisions, made by real AI.