Global scope · early development
We are building the AI expert for regenerative food production. Give it a location and a goal; it builds an understanding of the land, designs the production system, and adapts management as evidence and conditions change. The long-term control plane extends that intelligence into shared humanoid robots and small imaging drones.
Our future distribution direction centers on towns, villages and municipalities: a shared system serving community-registered land. The broader scope spans homes, smallholdings, market gardens, orchards, nurseries, farms and controlled environments, with biologically diverse regenerative production at its center.
01 · The first product
The first workflow begins as a founder-operated, AI-driven design service. AI is the domain expert: it enriches the site record, develops the plan, makes agronomic and management decisions, and adapts them as evidence changes. Deterministic tools ground geometry, measurements, constraints, and provenance.
Hosted frontier models let us build and validate this workflow now through a provider-neutral architecture. A domain-specialized Eden model is a mid-term direction, after the licensed evidence, verified outcomes, corrections, and evaluation benchmarks exist.
Property boundaries, survey imagery, terrain, soil, water, infrastructure, existing vegetation, climate, access, and owner or operator goals become one explicit site model.
A practical plan that can be inspected, edited, approved, implemented, and later used by the future control layer as the operational starting point for the site.
The customer sees a clear design. Underneath it, the system creates structured data that future sensing, automation, and robotics can operate against.
02 · How we work
Regenerative systems are site-specific and biologically complex. The AI is the expert; the site is the source of truth. It separates observation from inference, uses deterministic tools for computable facts, exposes uncertainty, and identifies any external approval that the applicable law actually requires.
Collect the minimum site data needed to understand real constraints before designing.
Represent land, assets, biology, risks, and goals in a versioned digital twin.
Reason across the whole production system and generate viable, site-specific options.
Check geometry, access, crop constraints, soil compatibility, and other deterministic rules.
Issue the best-supported plan, identify unresolved evidence, and separate design from regulated approval.
Record what was built and what happened so future recommendations improve with evidence.
03 · Real proving grounds
The sites are not three separate products. Each one stresses a different part of the same global architecture: site intelligence, biological planning, task execution, and outcome learning.
End-to-end product track
An accessible property with existing production areas and regular founder access.
Long-horizon observation track
Approximately 9,000 m² of hillside land with water and power.
Commercial discovery track
Labor-intensive commercial nursery and garden wholesaler.
04 · What we are proving now
The company vision includes autonomous stewardship and robotics. The immediate job is narrower: prove that AI can understand one site and produce a design that a qualified owner or operator can understand, trust, and actually use.
Given a site location and concise owner or operator goals, use AI to progressively enrich the site and produce a constraint-checked, machine-readable regenerative design with explicit assumptions, provenance, and uncertainty.
Site modelCanonical schemas and boundary ingestion
Design structureZones, crops, constraints, assumptions
AI expertStructured site and production planning
GA-001 packageStructured JSON plus readable design report
AI builds the site model and designs the production system.
GardenOS is the direction for observation, planning, work coordination and learning.
Humanoid robots, small imaging drones, people and irrigation serve participating parcels.
Outcomes improve decisions across sites and seasons.
Towns, villages and municipalities could buy or lease a shared solution. Community members would register eligible parcels; GardenOS would decide what to plant and where based on land suitability, community goals and access permissions. A shared fleet would carry out suitable planting, care and harvesting work.
Dedicated systems remain a secondary option for families with substantial land. Pricing, fleet economics, participation terms and produce allocation are still open.
Explore the humanoid and drone vision ↗05 · Early conversations
We are interested in speaking with municipalities, community organizers, landowners, growers, production operators, agronomic experts, geospatial specialists, robotics engineers, and potential research or pilot partners anywhere in the world.
This is an early-stage project. We are not yet offering autonomous gardening or guaranteed agronomic outcomes.