Food Business Canvas Method

Purpose

Use Karpathy-style LLM Wiki to turn the food-business search into a compounding map instead of one-off research reports.

The goal is not “find a food idea.” The goal is to build a structured idea radar that repeatedly asks:

Is this a real pain?
Who pays?
Can Sean prototype it quickly?
Does it avoid daily ops/perishables/food-safety hell?
Is there leverage from software, automation, sourcing, or local Nevada context?

Workflow

1. Bucket the opportunity space

Every idea gets placed into one or more buckets:

Picks-and-shovels food tech
Food infrastructure
Food distribution / logistics
Institutional / grant-backed food systems
Food waste / upcycling
CEA / greenhouse / mushroom automation
Consumer packaged goods
Kitchen / commissary automation
Procurement / sourcing intelligence
Cold-chain / food safety tooling

2. Create one page per idea

Each idea gets an `ideas/.md` page with:

One-line thesis
Customer / buyer
Pain
Product wedge
Why now
1–2 week prototype
Sean-fit score
Risks / why it might suck
Kill criteria
Related pages

3. Create one page per relevant company/entity

Companies are not just source links. Each company page should answer:

What do they sell?
Who buys it?
What is the pricing/positioning if visible?
What does this prove about demand?
What is missing/weak that Sean could exploit?

4. Create concept pages for repeating patterns

Examples:

Cold-chain monitoring
Food bank procurement
Mushroom-room automation
USDA grant reporting
China-sourced greenhouse hardware
Restaurant supplier pain
Food safety logging
Local food aggregation

5. Run comparison passes

After 10–20 idea pages, create comparison pages:

Best low-ops food businesses
Best Nevada-specific food businesses
Best picks-and-shovels ideas
Best 1-week prototypes
Ideas to kill
Ideas worth buyer interviews

6. Keep a kill list

Killed ideas are valuable. Mark them `rejected` with the reason so we do not keep rediscovering them.

Scoring rubric

Score 1–5:

Pain intensity
Buyer clarity
Prototype speed
Sean-fit
AI/automation leverage
Local Nevada/Reno/Tahoe edge
Margin potential
Low capex
Low regulatory burden
Low daily ops burden

High total is good, but any fatal score can kill the idea.

Fatal negatives:

Requires Sean to run daily perishables ops
Requires high capex before buyer proof
Commodity food margins
Heavy food-safety liability before revenue
Buyer is vague
Prototype takes months

First canvass targets

Start by making 20 fast idea pages across these themes:

1. Mushroom-room controls kit

2. Greenhouse automation retrofit

3. Cold-chain monitoring for small producers

4. Food bank procurement/reporting dashboard

5. Restaurant supplier price/availability intelligence

6. China CEA hardware sourcing database

7. Modular solar cold storage

8. USDA/local food grant reporting SaaS

9. Local producer aggregation marketplace

10. Food safety logbook app for small producers

11. Kitchen prep automation for commissaries

12. AI menu/procurement optimizer for meal prep businesses

13. Food waste analytics for restaurants

14. Upcycled ingredient microfactory

15. Freeze-dried local food products

16. Shelf-stable sauce/soup production system

17. School greenhouse monitoring kit

18. Tribal food sovereignty tech stack

19. Tahoe/Reno local food discovery/distribution platform

20. Commercial kitchen pod / permit-ready microfactory

Output cadence

Each research sprint should produce:

5–10 new/updated pages
1 comparison page
3 ideas promoted to “interview buyers”
3 ideas killed with reasons
next source list

Why this is better than deep research reports

A normal report collapses the space too early. This wiki preserves weak signals, adjacent examples, killed ideas, and cross-links. After enough pages, the best ideas should emerge from comparison rather than from a single model’s vibes.