Structural read, not agronomy or investment advice. The author is applying the 10-layer framework, not a domain expert in farm machinery or precision agriculture.
The setup. John Deere's See & Spray uses vision models to identify weeds in real time and selectively spray only those plants, cutting herbicide use by a large margin. It is one of the clearest examples of "AI in the physical world" actually shipping at industrial scale.
Through the layers. • L-1, physical fleet & edge compute. Cameras on the spray boom, GPUs in the cab, the tractor itself. Deere has decades of fleet density and a dealer network that finances the install base. This is the layer that does not commoditize when a new foundation model ships. • L1, agronomic data. Every pass of every machine generates weed maps, yield data, soil response. Tied to specific fields, specific farmers, multi-season. A new entrant cannot scrape this. • L3, vision models. Required, but not the moat. The model is the layer where the competition lives, not where the value sits. • L8, workflow & financing. The integration into planting, spraying, harvesting decisions, plus Deere Financial, is what makes the customer renew. Software-only entrants don't own this.
Why this case matters for the framework. Most AI analysis treats the model as the center of gravity. In physical-world AI, the model is the easy layer. The hard layers, L-1 (the physical asset and the edge silicon inside it) and L8 (the operating workflow and capital structure around it), are where the durable value sits. A foundation model from any frontier lab cannot run See & Spray without a Deere tractor underneath it.
How the layers behave differently here. • L-1 takes years and capital to build. Software companies treat L-1 as a rounding error. In agriculture, robotics, energy, and manufacturing, L-1 is the dominant layer. • Cycles are slower. Farmers don't refresh tractors yearly. Layer compression that takes 18 months in software takes 5–10 years here. • L8 includes financing. In SaaS, L8 is the workflow. In industrial, L8 also includes the multi-year financing structure that locks in the install base.
Worth watching. Whether a software-native entrant (e.g. Carbon Robotics, or a startup pairing with Kubota or AGCO) can assemble enough L-1 to start eating Deere's data flywheel, and whether open vision models commoditize L3 fast enough to make the L-1 advantage less defensible over a decade horizon.
Public reporting from Deere investor materials; rollout figures approximate as of May 2026.
- What This Means for You
Product Leader
Map your product to the layers it actually owns vs. rents. The rented ones are where the counter-move work belongs.
Investor
Underwrite layer ownership, not feature count. The Cube footprint is the moat.
Operator
Audit your stack against Supply Chain of Intelligence. Anything sitting only at L7 is the layer to watch.
AA
Anand Arivukkarasu
Ex-Meta product leader. Creator of Supply Chain of Intelligence™. Writes about where AI value accrues, and who can fire your product. LinkedIn