@#*+@=:-:*:%@: @+@.@+##@-=@*: %==. .-#+%#.*. +:@.@=%=*:+=-*
Loading
The systems that fed the world for the last 100 years are collapsing
In order to feed billions, we've oversimplified food production by growing the same crop on thousands of acres. The existing farming equipment, supply chains, and farmers' knowledge are optimized for single crops. We assume that plants are isolated systems that simply need water, light, soil, nutrients, and air to grow.
But that couldn't be further away from the truth. A single acre of rainforest contains over 500 plant species that form ecosystems which have evolved for centuries to coexist and support one another.
As climate change worsens, pests become more frequent, and global instability rises. The fragility of monocultures is becoming increasingly apparent. When one variable is wrong enough to kill the single crop in the system, everything goes to zero.
As of today, 2.3 billion people don't have reliable access to food, and it's become clear that the only way we have to feed the world in the coming decades is to truly understand biology and use it to our advantage.
The Master Plan
Imagine a world where an Agroecological Foundation Model designs custom biological ecosystems for any plot of land on Earth based on its conditions and climate projections. These ecosystems would double food production per acre while making it resilient to the changing climate.
Cracking plant biology is hard. The reason billions have been spent and lost in this field is that translating findings into the real world is absurdly difficult. There are countless variables in every ecosystem that are impossible to test or account for in isolation.
We believe in leveraging some of the 500 million non-industrialized smallholder farmers to create a global, decentralized reinforcement learning network that trains the Agroecological Foundation Model in real ecosystems without even realizing it.
Phase 1 — Win-Win
On average, smallholder farms lose 30% post-harvest due to a lack of infrastructure. We deploy containerized processing units that transform raw harvests into shelf-stable commodities. Each unit is equipped with sensors that collect unparalleled data on soil, weather, and nutritional output.
Phase 2 — Biological Recipes
A digital twin is created based on real-world data, and the Agroecological Foundation Model is trained to design custom polyculture recipes for farms in the network. Neighboring farms remain as controls while the units process the outputs of both farms and feed the data back to the model, creating a positive flywheel.
Phase 3 — Terraforming
The Agroecological Foundation Model has been proven effective for years on small-scale farms and is ready to transform industrial farming by designing custom biological ecosystems at scale and managing modular robots that handle the complexity of growing resilient food for 10 billion people in a changing climate.
Needless to say, this is a moonshot project. Achieving this in a decade will take billions of dollars, some of the greatest minds in machine learning, hardware, and biology, and we may still fail.
But if we succeed, we change the course of humanity.
Half of the world population will not have reliable access to food by 2050
Monocultures are optimized for scale but fragile by design
A single climatic event can disrupt entire food supply chains globally