A network of local farms and kitchens, coordinated around the next meal. Ingredients for people and future home humanoid cooks. Prepared meals from a neighborhood kitchen. Software connecting the work, with robotics taking on useful physical tasks as capabilities mature.
Starting in Bozeman. The integrated system is in development.
Conceptual future network. Household ordering, prepared-meal delivery, and home-robot integration are not currently operating Streamline services.
The investment thesis
Local food infrastructure for a world of physical AI.
Today, the farm, the kitchen, and the household plan separately. Streamline's thesis is to connect them into one operating system: demand informs what gets grown, available food informs what gets made, and every physical task has a cost and a clear place in the plan.
As machines become capable of more growing, handling, and cooking tasks, food production could move closer to the people it serves. We're developing the local infrastructure those capabilities would need. One neighborhood becomes the proving ground; the larger asset is a food operating model repeated across sites and operators.
The integrated operating model is the direction we are developing. Current software reconstructs recorded greenhouse evidence; it does not yet coordinate that complete network.
A crop can grow well and still miss the meal. Growing, harvesting, packing, preparing, and delivery often run on different schedules. A neighborhood farm and kitchen should connect those jobs around the people who will eat the food.
A conventional route
Illustrative
Grow
Farm
Produce and harvest food.
Prepare supply
Aggregation, processing & storage
Combine supply, prepare products, and hold inventory.
Move
Distribution
Transport food to the next buyer.
Offer
Retail or food service
Sell ingredients or prepare a meal.
Eat
Household
Choose, cook, or enjoy the meal.
Actual routes vary: businesses can combine steps or sell directly. Aggregation, processing, storage, distribution, and retail perform useful work; this diagram does not treat those functions as waste.
The neighborhood route we want to test
Proposed · In development
Grow + source
Neighborhood growers + partner suppliers
Plan suitable local crops and source the ingredients the neighborhood cannot grow.
Prepare + coordinate
Kitchen & packing
Portion ingredients for home cooking or prepare meals, then coordinate the household handoff.
Cook or enjoy
Household
Receive ingredients for the next meal, or a prepared daily lunch or meal.
This integrated service is not operating yet. Partner ingredients retain their upstream supply chains. Inventory checks and acceptance, food handling, cold storage, preparation, and delivery still have to be provided and paid for.
The test is the delivered meal. Measure handoffs, lead time, waste, and delivered cost per accepted meal. These are illustrative process maps, not measured supply-chain data or a promise of fewer handoffs, lower costs, or less waste.
The proposed network
A farm. A kitchen. A neighborhood. One connected plan.
01 · PRODUCTION
A neighborhood production node
Local crops and partner supply would connect to dependable demand. Growing, sensing, and selective handling create bounded tasks where useful automation can be tested.
02 · PREPARATION
A kitchen that connects the system
Packing ingredients and preparing meals would share a view of available food, quantities, handling, substitutions, and delivery. This is physical infrastructure as well as software.
03 · THE NEXT MEAL
People, kitchens, future home cooks
A household could cook the ingredients, receive a prepared daily lunch or meal, or eventually use a qualified home humanoid. Each path needs its own cost and operating requirements.
The integrated network is in development. A pilot must establish the offer, safe food operations, supply, pricing, and fulfillment. Household ordering and meal delivery are not open yet.
Future research · Home cooking
Ingredients for the person cooking. And, one day, a home humanoid cook.
A meal still starts with food, whoever prepares it. We want to explore how a neighborhood farm and kitchen could supply ingredients and preparation information to people today and future home humanoids as their cooking capabilities develop.
The right ingredients
Recipe quantities, portions, available lots, and approved substitutions would need to agree before ingredients leave the kitchen.
A useful preparation record
Provenance, storage and handling information, ingredient state, and preparation steps would need to travel with the food.
A confirmed handoff
A household or future robot cook would need to confirm what arrived and stop for missing ingredients, uncertain instructions, or an unapproved change.
This is a research direction. Streamline has no verified home-robot cooking integration or compatibility today. Each device and workflow would need testing, human approval, and clear responsibility for food handling. The first household service would not depend on owning a robot.
Put more of the food dollar to work close to home.
Growing food is one part of what a household pays for. The bill also covers preparing it, storing it, moving it, and getting it onto the table.
Our opportunity is to do more of that useful work in one neighborhood: grow suitable crops, bring in partner ingredients, prepare meals, and serve nearby households. Earning more of the purchase also means paying for the work behind it.
went to farms for commodities used in domestically produced food purchased by U.S. consumers, across grocery and food-service spending. This is gross farm receipts before farm costs, not farmer profit or a measured Streamline share.
The test is what remains after ingredients, people, kitchen work, delivery, and infrastructure. Our four-acre scenario is a planning example, not an operating result or a promise of returns.
Bennett Cawthon began Streamline’s year-round hydroponic produce operation in Bozeman in 2017. The experience tied crop choices to harvesting labor, seasonality, and customers.
Current engineering
A recorded evidence foundation
Streamline OS reconstructs a bounded recorded greenhouse interval. An early-access record-analysis tool is available. These establish engineering work, not an operating neighborhood food service.
Next commercial proof
A paid household pilot
The proposed pilot must demonstrate renewal, usable crop output, reliable fulfillment, contribution per order, and complete site economics. Its customer count and returns are not established.
Build the service now. Prove better tools as we go.
The first useful meal does not have to wait for a fully autonomous farm. Start with people, proven equipment, and a clear household offer. Test new tools against the actual work, then adopt them where they improve cost, consistency, or capacity.
Bounded recorded greenhouse reconstruction and an early-access record-analysis tool. These support investigation; they do not operate a household food service.
NEXT02
Build the first integrated proof
Capital and operating partners would connect a priced offer, production and supply, kitchen capacity, and fulfillment. Measure accepted food, paid renewal, exceptions, and complete contribution.
LATER03
Qualify better tools
Compare automation with the actual task baseline. Adopt it only where safety, useful output, supervision, and complete cost justify it. Repeat the service only after its economics work.
Earlier food-farm economics pushed Streamline toward a higher-value crop and deeper engineering. Returning to food still requires a viable offer, paid labor, safe handling, and complete capital costs. A more complete technology stack would not, by itself, prove that business.
Saleable crop
Must carry: full harvest cost
Finished meal
Must carry: full fulfillment cost
Completed robot task
Must carry: supervision and ownership cost
In Streamline's earlier tomato operation, too many picking actions for a saleable pound made harvest labor a practical constraint. See the snips-per-pound example.
The larger goal is a repeatable operating model. Another site and operator would need to validate it after the first neighborhood establishes full costs and daily responsibilities.
For robotics teams and builders
Bring one useful capability to the neighborhood.
Help investigate a real task between crop and kitchen. Start with a baseline, define acceptable work, and show the help it needed and what the whole job cost.
01
Observe the crop
Repeatable scouting or capture for operator review. Measure coverage, useful findings, missed observations, and review time.
02
Move materials
A defined transfer of compatible trays or crates at an assessed site. Measure time, damage, interventions, and cleaning.
03
Handle produce
One crop and one picking, sorting, or packing task. Define accepted output, then measure rejects, supervision, and cost per accepted unit.
04
Verify the handoff
Reconcile an item, quantity, and order between farm and kitchen. Measure errors and resolution effort. A scan alone does not establish food safety.
05
Future home cooking
Later exploration
Explore ingredient and recipe information for a home humanoid. Physical handling and cooking require separate compatibility, qualification, and safety evidence.
Human review stays alongside the task.
What job?
What counts as acceptable?
What help did it need?
What did the whole job cost?
Tell us what your system demonstrates today, what conditions it needs, and how you would compare it with the existing method. We are looking for useful work another operator could repeat.
Exploratory collaboration inquiries. These are candidate studies, not existing deployments or a paid testing offer. Test sites, scope, access, and commercial terms would need to be agreed separately.
The investment and building thesis for a future network of neighborhood farms and kitchens: useful ingredients, prepared meals, and qualified home-cook integration.
An illustrative farm-and-kitchen ledger tests what household demand must cover: ingredients, preparation, neighborhood delivery, overhead, and capital.
A synthetic service-design walkthrough tests one future network node: household permissions, crop planning, ingredient handoffs, kitchen preparation, and delivery.
agricultural sensingfarm datafood logistics
Why handling matters
Freshness takes more than a short distance.
Crop, storage time, temperature, and preparation all matter. A nearby farm still needs careful cooling and handling, and a kitchen needs a record of what arrived.
One crop, one study: packaged spinach
A 2004 storage experiment reported 53% of initial folate remaining at these different time-and-temperature endpoints.
4°C
8 storage days
53% of initial folate retained
10°C
6 storage days
53% of initial folate retained
20°C
4 storage days
53% of initial folate retained
The spinach was visually unacceptable at these endpoints. These are experimental storage durations, not known days since harvest, safety guidance, or a universal nutrient-loss curve. They do not measure Streamline food or establish nutritional superiority for local food.Pandrangi & LaBorde, Journal of Food Science (2004) · Study record and DOI
For the proposed pilot, the useful test is crop-specific: log harvest and receipt times, storage conditions, accepted quality, and waste through the household handoff. Any claim about nutrient retention would need its own sampling and measurement.
From food, through cannabis, back to food
The experience shaped the company. Dinner is the destination.
Bennett Cawthon began as a food farmer. When food-production economics pushed the operation into cannabis, the higher-value crop supported deeper development of farm software, sensing, spatial systems, and robotics.
Regulated cultivation also made a deeper question economically legible: how environment, crop work, and time shape genetic expression. Streamline is bringing those capabilities back to food and building the evidence needed to test that thesis without mistaking correlation for causation.
Help build the infrastructure behind the next meal.
The first integrated neighborhood can reveal what the next one needs. We're seeking aligned capital and partners in growing, kitchens, supply, land, software, and robotics to build that proof together.
A first conversation should identify the contribution, milestone, and evidence that justify the next step. Investment terms and collaboration scope would be agreed separately.