Ten Food-System Assumptions Worth Testing

A forecast is useful when it changes a decision for a reason that can be checked.

The food system offers plenty of dramatic predictions: robots replacing labor, neighborhoods becoming farms, kitchens becoming optional, and data becoming a new agricultural asset. The harder work is identifying what would have to become true before investing around those possibilities.

These are ten assumptions Streamline would test. Some describe genuine constraints; others describe habits or business arrangements that might change. None comes with an expiration date.

1. Routine meals will continue to require the same amount of household work. Automated preparation and delivery could change the balance between cooking at home and purchasing prepared food. The test is a service people choose repeatedly at a price that covers its full cost. Cooking also has value beyond the meal: control, culture, enjoyment, and time with other people.

2. One large production site will always beat many smaller ones. Concentration makes supervision, equipment, packing, and logistics easier. Smaller sites would need enough advantage from proximity, existing resources, or crop fit to cover their extra coordination. Existing food hubs already connect multiple producers and buyers; the unproven extension is cheaper qualified work across more fragmented sites. USDA local-food glossary

3. Agricultural machines will need to earn their return from one task or site. Shared machinery and contracted services already offer alternatives to individual ownership. More adaptable robots could extend them if switching tasks becomes economical. Look for paid work across compatible jobs, with transport, setup, supervision, downtime, and ownership costs included.

4. The task model will remain the most defensible part of a robotics business. Reusable policies and demonstrations could make some capabilities easier to reproduce. That would increase the importance of reliable integration, qualification, service, and customer relationships. The relevant evidence is the cost of reproducing a dependable deployment, not a similar demonstration.

5. More automated work means a better operation. A machine can complete a job while increasing rework or moving a bottleneck elsewhere. The test is improvement in the complete workflow against a credible existing alternative. Farm unit economics supplies the comparison.

6. An interface labeled AI provides better decisions. A dashboard can be valuable, and sophisticated analysis can appear inside one. Product labels settle little. Ask whether using the system improves a defined decision, whether the evidence is inspectable, and whether that result survives a prospective comparison. Agricultural intelligence should be judged by that standard.

7. Collecting more agricultural data necessarily creates more value. A large archive can be unusable if measurements lack identity, timing, rights, or comparable outcomes. A smaller record may support an important decision. The test is whether the additional data improve a validated use enough to justify collection, maintenance, and governance.

8. One favorable resource metric establishes sustainability. Lower water use, shorter travel, and higher yield each reveal part of a system. They can coexist with increased energy, packaging, or material use elsewhere. Require a declared boundary, comparable output, and the material tradeoffs. NRCS's use of multiple indicators to assess soil function offers a useful example of why one measurement cannot settle a broader claim. NRCS soil-health assessment

9. Year-round availability will sustain demand for the same product. Our Bozeman experience with same-day fresh tomatoes showed the limit of that assumption: customers still changed what they wanted to eat in winter. Test repeat purchasing, seasonal preferences, and commitments separately from production capability. A successful growing system can serve a seasonal market.

10. Growth establishes that the underlying business works. More volume can spread overhead and improve purchasing. It can also increase losses and capital requirements. Look for a repeatable unit with acceptable returns, then measure whether expansion preserves or improves them.

Each assumption should connect to a practical choice: equipment purchased, capacity reserved, crops planted, software developed, or a service offered.

Write down the evidence supporting the choice, the observation that would change it, and the smallest useful test. Account for evidence against the thesis as carefully as evidence in its favor.

That is also how the backyard-network scenario and the proposed World Biological Outcome Model should be read. They describe opportunities whose dependencies can be investigated.

Conviction becomes more useful when it names the evidence that could change the plan.