# Anyground — a platform for places that keep working

**Project vision and platform specification · research-stage proposal · 29 October 2026**

> Start with a useful capability. Connect it to a place. Keep it working. Let the place change.

## Reading this document

Anyground is a proposed system for creating, operating, maintaining and expanding complete living environments. It is not a released hardware catalog, a construction specification, an offer to sell housing, or a demonstration of autonomous self-sufficiency. All Anyground products, interfaces, service targets and roadmap gates below are proposals unless explicitly identified as measured evidence. Numbers labeled illustrative are assumptions for reasoning, not equipment ratings or promises.

The supplied 98-page *Vision-and-Strategy.pdf*, dated 26 October 2026, already contains a sophisticated platform vision, technical review and visual atlas under the working name Commonbeam. This document preserves its strongest decisions, develops missing operational detail and uses **Anyground**, the name requested for this project. Neither name has received legal clearance through this work. Internal architecture study names are not additional public brands. The separate critical review records where the source should be retained, challenged or narrowed.

## 1. The complete ambition

Anyground should make establishing a useful place feel more like choosing from a coherent catalog and less like managing an accidental federation of incompatible contractors. A customer describes what people need to do, where the place stands and what support is available. The platform translates that need into a configuration: products, approved alternatives, resources, site prerequisites, installation responsibilities, operating limits and a repair plan. The result is a place with understandable obligations, not merely a shipment of attractive objects.

The long-term scope includes ordinary homes, productive gardens, workshops, remote lodges, field stations and eventually components for extraterrestrial habitats. The ambition is large because the same lifecycle problems recur across them: equipment must connect, resources must balance, people need comfort, failures must remain containable, and useful parts should survive a change of use. The hardware need not be identical for the platform to make those jobs substantially easier.

A complete environment includes more than shelter. It must support rest, food preparation, hygiene, privacy, social life, work, movement, storage and access. It needs power, water, temperature and humidity control, sanitation, waste routes and maintenance. It also needs permissions, an accountable operator, resupply, skills, spare parts and a way to leave the site responsibly. These are product requirements, not externalities that can disappear behind the word sustainable.

The proposed initial product is **a bounded, staffed growing capability on known utilities**, supported by a common service architecture. It is one proving ground, not a redefinition of the company as a vertical farm. Core + Grow + Climate + Care can test identity, isolation, metering, repeatable work and paid support before occupied structures or remote energy systems multiply the hazards. The buyer choice remains a commercial hypothesis: an existing grower and a reachable professional kitchen must compete on observed need and willingness to pay.

The platform should accept a useful retreat. If a complete growing installation is commercially weak but a serviceable irrigation-and-handling module is valuable, sell that bounded capability. If agriculture proves a poor entry market altogether, do not preserve it merely because it produced compelling images. A resilient domestic utility core or workshop retrofit is an alternative only after equivalent customer evidence, not another simultaneous launch.

## 2. IKEA and MUJI: operating disciplines, not a visual costume

IKEA's democratic-design precedent considers form, function, quality, sustainability and low price together. Its broader relevance is coordinated product, packaging and production design, rather than an assumption that a new venture automatically inherits global purchasing power. [IKEA Museum guide](https://ikeamuseum.com/wp-content/uploads/2025/09/democratic-design-schools-university-teacher-guide_250331_gb.pdf)

MUJI describes material selection, process streamlining and packaging simplification as its underlying principles. Importantly, its account does not reduce this to an aesthetic style. Anyground should learn from that distinction: natural-colored surfaces are not evidence that a system is economical or materially responsible. [MUJI's own account](https://www.muji.com/jp/about/en/)

For Anyground, the combined discipline means:

1. **Design the delivered service and its whole chain together.** A beautiful water drawer that requires a bespoke technician visit every week is a poor product, even if the drawer is inexpensive to fabricate.
2. **Reduce total decisions, not merely visible controls.** A configuration should include appropriate valves, wiring, spares and work instructions rather than force the buyer to reconcile hidden incompatibilities.
3. **Minimize variation where it reduces real work.** A shared service-drawer envelope may be useful; forcing a bedroom, crop tray and pressure vessel onto one arbitrary grid is not.
4. **Keep essentials available without decorative complexity.** Local controls, repair access, a legible fault state and ordinary tools take priority over a fleet dashboard.
5. **Make the ordinary entry product desirable.** Modesty in size and capability is compatible with excellent daylight, tactile materials, quiet operation and precise detail.
6. **Distinguish installation speed from readiness.** A prepared assembly can be rapid. Site assessment, drainage, approvals, commissioning, occupancy and biological growth still have their own timelines.

The one-day ideal is a useful *experience target for a prequalified, prepared configuration*, not a universal construction promise. Track the entire elapsed customer journey, including sitework and waiting for specialists. Optimizing only the visible erection event would repeat the fragmentation Anyground intends to remove.

## 3. Universality in three honest layers

A universal platform is not a universal object. Every reuse statement should name its level:

| Reuse level | What can remain common | What must still be proved |
| --- | --- | --- |
| Exact part | An identical crop carrier, bracket, sensor module or service tool | Loads, fluids, temperature, hygiene, electrical rating, wear and firmware compatibility in the destination |
| Qualified family | The same interface envelope, handling method and repair sequence with different seals, coatings or internals | Each variant's performance, permitted substitutes and family-level assumptions |
| Capability contract | Different devices describe the same useful service in the same units, states and evidence format | Whether the specific configuration delivers the required service under its actual conditions |

A terrestrial water pump might repeat unchanged in two temperate grow installations. A saline-water pump might share a drawer, tools and diagnostic vocabulary but require different wetted materials. A lunar fluid unit may share only the identity, task and service model. Showing one silhouette in all three places must not imply common certification or common internals.

The original source correctly resists choosing a fashionable universal geometry first. Retain that restraint. Select working dimensions from transport routes, ergonomics, crop workflows, purchased equipment and load studies. The website's illustrative geometry is not a dimensional release. A standard can be developed only after the tolerances, installation access and tests are understood.

## 4. The architecture: support, service and independent cells

The proposed architecture has six layers:

1. **Site layer:** survey, access, ground support, water source, drainage, exposure, hazard assessment and local permissions.
2. **Support and envelope:** enduring frame, bracing, weather and fire strategy, insulation, access, daylight and optional mission-specific pressure boundary.
3. **Service Spine:** accessible, segregated utility routes; branch isolation; metering; local control; component identity; test and service access.
4. **Capability cells:** living rooms, kitchens, washrooms, growing bays, storage and workshops. Each states what it needs and what it delivers.
5. **Shared resource cells:** electrical conversion/storage, thermal services, water treatment/storage and managed waste. Sharing is optional and bounded.
6. **Assurance and Care:** configuration rules, commissioning, operating history, diagnostics, inventory, training, changes and decommissioning.

**The Service Spine is the principal proposed integration object.** It is not one magic multi-fluid plug. It is an arrangement of distinguishable connection families and repeatable work: find the branch, identify it, isolate it, verify absence of hazardous energy or flow, replace the correct module, test it and restore the service. The same procedure should remain recognizable as a site grows.

Use a topology of independently isolatable cells. A failed nutrient loop must not infect potable water. A battery fault must not share an occupied cupboard with the only exit. A single software controller must not be the sole authority for every protective action. Common infrastructure can reduce cost while still creating common-cause failure; the layout and control system must expose that trade-off explicitly.

The architectural baseline is the supplied **Living Rack** idea: durable support with accessible services and replaceable infill. **Pleated Ground** remains an expressive envelope investigation. A fixed folded skin can retain rhythm, shading and stiffness without moving joints. A secondary deployable canopy can be investigated without making the occupied weather boundary deployable. A moving pressure boundary is a much more demanding, separate research problem.

## 5. Catalog logic

The catalog should be organized around functions people can understand. The following families are a development structure, not eight simultaneous product launches.

| Family | Function | Initial treatment | Later development |
| --- | --- | --- | --- |
| Core | Identity, local control, metering, isolation and service organization | Owned integration layer, small approved device set | Common local coordinator and configuration records |
| Grow | Cultivation, predictable carriers, irrigation and crop work | One crop family and one complete workflow | Additional crop/handling modules after proof |
| Climate | Shade, ventilation, heating/cooling, humidity and envelope interfaces | Qualify established equipment for the pilot | Regional packages with seasonal evidence |
| Care | Repair, spares, training, maintenance and accountable support | Mandatory operating plan, optional managed work | Regional refurbishment and selected robotic exchange |
| Water | Intake, treatment, storage, circulation and recovery | Existing known supply with qualified partners | Source-specific packages and tested reuse loops |
| Power | Generation, protection, distribution, conversion and reserve | Known electrical supply and declared outage response | Microgrid options for qualified loads and sites |
| Flow | Material movement, handling, washing, packing and cold storage | Count all tasks even when performed by people | Dedicated products where repeated demand supports them |
| Home | Habitable spaces and domestic functions | Later family; professionally supplied enclosure if needed | Qualified rooms, kitchens, sanitation and complete places |

Frontier is an internal research program for difficult environments, deployment and advanced autonomy. It is not a consumer range. A frontier component enters the catalog only after it has a useful, supported scope.

Every catalog page should answer: what job does this do; what is included; what must the site provide; how does it connect; what is its operating domain; what remains human work; what fails; how is it repaired; what does it cost to own; what evidence exists; and what is specifically excluded? Product naming should never carry a capacity claim whose measurement is undefined.

## 6. A module record is part of the physical product

Each module requires a durable ID, version and serial history. Its public record should separate design targets, measured values and supplier statements. At minimum:

- physical envelope, mass, center-of-mass/handling restrictions and service clearances;
- mounting family, load class and applicable installation orientation;
- typed electrical/fluid/data ports, ratings, materials and permitted consumables;
- operating inputs and outputs, including quality, units, duration and environmental conditions;
- dependencies and resource peaks, not just averages;
- local modes, protective behavior and allowed task sequences;
- cleaning, calibration, inspection, wear and replacement methods;
- approved replacement parts and incompatible revisions;
- software support, access, export, backup and recovery provisions;
- commissioning evidence, inspection signatures and configuration-specific limits;
- end-of-life, contaminated-return and refurbishment routes.

An illustrative information record—not an implemented schema or certified product—could be:

```json
{
  "module_id": "example-irrigation-drawer",
  "revision": "concept-01",
  "maturity": "proposed",
  "capability": "metered_nutrient_circulation",
  "ports": [
    {"type": "nutrient_supply", "rating_ref": "not-yet-qualified"},
    {"type": "nutrient_return", "rating_ref": "not-yet-qualified"},
    {"type": "protected_electrical_supply", "rating_ref": "not-yet-qualified"}
  ],
  "required_tests": ["identity", "isolation", "leak", "flow", "alarm", "restart"],
  "failure_action": "isolate_branch_and_alert_named_operator",
  "supported_configurations": [],
  "evidence": []
}
```

A record with empty qualification evidence must not be returned by a commercial configurator as a supported purchase. Conversely, an attractive animation is not evidence that a module record is complete. The digital model should constrain promises rather than decorate them.

## 7. Interfaces: make the wrong connection difficult

Mechanical attachment, utilities, handling and software need distinct contracts. A shared service arrangement can place them coherently without pretending they are interchangeable.

| Interface | Required design decisions | Acceptance principle |
| --- | --- | --- |
| Structural | Load path, tolerances, inspection access, uplift/lateral restraint, thermal movement | Connection is verified for actual loads and substrate, not merely assembled |
| Potable water | Approved wetted materials, backflow strategy, drainability, cleaning and sampling | Correct use and quality are established at the service outlet |
| Nutrient water | Chemistry, biofilm control, crop isolation, purge and sanitation | Measured circulation and containment; no unverified domestic crossover |
| Waste | Gas/odor management, venting, accessible cleanout, containment | No uncontrolled release; maintenance has a defined destination |
| Thermal fluid/air | Condensation, pressure, flow, insulation, service disconnects | Delivered thermal service at required temperatures and conditions |
| Power | Voltage/current, protection, bonding, isolation, fault behavior, local requirements | Qualified specialist commissioning with appropriate evidence |
| Data | Identity, units, timestamps, command authority, stale-sensor behavior | Physical outcome is independently verified where the hazard requires it |
| Robotic handling | Grasp surfaces, alignment, payload, locked states and approach clearance | Success includes seating, verification and recovery, not arm motion alone |

Color is helpful but insufficient for type distinction. Use different geometry, keying, labels and documented tests. A data handshake cannot make an inappropriate seal or conductor safe. An adapter must carry its own controlled compatibility record rather than become a universal bypass.

CfOC/ICC 1220 is relevant existing work on off-site connections, not evidence that Anyground has invented modular interoperability. The committee's public page describes an in-development standard. An October 2026 update reports initial physical interface, clearance and access studies. Anyground should follow its current status and use applicable standards without claiming draft work is a universal approval. [Committee](https://centerforoffsiteconstruction.org/modular-consensus-committees/), [prototype update](https://blogs.nyit.edu/build_ahead/cfoc_icc_1220_project_prototype_update_office_conversion_installation_schematic)

## 8. Living space: comfort is an essential system

A technically efficient habitat can still be a poor place to live. The Home program should establish requirements for daylight, view, glare, acoustic privacy, sleep, usable storage, ventilation, moisture, cleaning, social space and access before selecting a dramatic exterior. A person should not need the robotic system to succeed to open a door, reach a manual control or leave during a fault.

Japanese influence should act as a design method. An engawa-like threshold can mediate interior, climate and shared space; it is not appropriate as an unprotected winter circulation route in every climate. Repeated proportions can make infill legible without forcing all functions to identical dimensions. Sliding partitions can allow daily change if acoustic privacy, fire and access remain addressed. Precise joints should reveal how parts are serviced, not fetishize concealed complexity.

Keep hazardous, noisy and dirty systems outside ordinary living zones or behind appropriate separation. The household should have a calm resident-facing interface, while trained maintainers receive the detailed service state. A living room can remain warm and tactile even if the service side is industrial. That duality is a stronger common language than making the whole place look like a machine room.

The first Home configurations should be assessed with users across ages and abilities. Reach distances, turning space, slip risks, wet-area access, changing daylight and sound matter. A transformable room must still work for someone who cannot move a heavy panel. Human agency includes the option to keep a simple fixed arrangement.

## 9. Energy and climate: start with the difficult interval

The order is demand reduction, passive response, efficient active services, then generation and reserve. A solar canopy is a component, not an independence proof. Model chronological weather and loads, peak demand, battery state, cold/heat limits, aging, maintenance and backup failure. Distinguish grid-connected efficiency, short-outage resilience and year-round remote operation as different products.

Each configuration needs a critical-load policy. Protective functions, required ventilation, freeze prevention and essential circulation may take priority over supplemental crop lighting, comfort peaks or other flexible work. But biological loads are not infinitely deferrable: a long lighting reduction affects production, and heat stress can destroy the next harvest. Record the cost of degradation as well as the electrical savings.

Use different forms of storage for different jobs. Batteries support electrical work; well-chosen thermal storage may serve heat or cooling more directly. Heat recovery only has value when demand exists at a compatible temperature and time. Do not credit lamp electricity as both useful crop light and free recovered heat without comparing the complete system with the separate services it replaces.

A greenhouse offers sunlight but loses moisture and may need ventilation. A sealed farm recovers more moisture but requires dehumidification and heat rejection. Evaporative cooling trades electricity against water. The platform must expose these couplings so a water-saving option does not silently create an unaffordable energy load. Climate packages should be qualified by location and season, not named only “desert” or “polar.”

## 10. Water and sanitation: qualify the service, adapt the source

Standardize the useful outlet and its evidence; vary acquisition and treatment by site. Mains, wells, surface water, rain, treated reuse, desalination and delivered water are legitimate alternatives when permitted and properly engineered. Atmospheric harvesting is an optional research candidate, not a mandatory foundation of Anyground.

Keep four different claims separate: collected volume, treated usable volume, water suitable for a particular use, and reliably delivered service over a season. Likewise, recovery fraction is not independence: losses, purge, contamination, equipment downtime and reserve filling still demand supply. Measure domestic and crop loops separately. Never count captured crop condensate as both new water production and recycled irrigation.

NASA's ISS account reports a 98% recovery milestone in a specific life-support configuration. It supports the possibility of high recovery with substantial engineering; it does not establish an inexpensive terrestrial appliance or remove replacement and resupply needs. [NASA water-recovery milestone](https://www.nasa.gov/missions/station/iss-research/nasa-achieves-water-recovery-milestone-on-international-space-station/)

Sanitation is not a generic recycling icon. Specify toilet and waste routes, containment, venting, access, treatment, outgoing solids/liquids and the person responsible. Dry toilets, conventional sewer, holding tanks, packaged treatment and nutrient recovery may suit different sites. None automatically licenses direct food-crop use. Nutrient recovery must track pathogens, trace contaminants, salt balance and actual crop requirements; a nutrient that is present is not necessarily available, safe or in the right proportion.

Provide finite reserves, sampling/inspection arrangements, consumables and a safe response to suspected contamination. A loss of water-quality confidence should restrict service appropriately even when a tank remains full. Automated sensing supports the plan but cannot substitute for every required quality test.

## 11. Food: produce useful output, not a self-sufficiency image

The early Grow offer should deliver a specified crop basket to an identified use. Fresh mass, plant count, calories, protein and micronutrients are different measures. A beautiful salad harvest does not demonstrate a complete diet. Broad food independence also requires seed, nutrients, substrates, pest management, processing, storage, cooking, losses and human dietary suitability.

Begin with predictable geometry and one crop family, selected with an experienced grower and a real buyer. Prefer daylight or an existing enclosure if it wins the local comparison. Fully artificial lighting is valid where isolation, constraints or crop value justify it; it should not be a visual default. A planning canopy area is not a product capacity until demand and agronomy establish it.

Map the whole workflow: receive inputs, sow, establish, irrigate, monitor, manage pests, move, harvest, wash if appropriate, pack, store, clean, reset and repair. Include rejected starts, unsaleable yield, crop rescue, disposal and remote expert time. Automation begins with fixtures, metering and dependable local controls. A gantry or conveyor earns its place through measured labor and reliability benefit, not because a generalized robot looks futuristic.

EDEN ISS is a useful extreme-environment experiment, not a demonstration of no-work food independence. A DLR thesis separates on-site and remote work and reports different output across experiment years. The lesson for Anyground is methodological: count both locations' labor and the variation between cycles. [DLR crew-time investigation](https://elib.dlr.de/210230/1/2024_PhD_Augmented%20Reality%20in%20a%20Planetary%20Greenhouse%20for%20Crew%20Time%20Optimization_Zeidler.pdf)

## 12. Autonomy means completing the recovery work

Use a ladder of bounded capability, not one autonomous label:

1. **Instrumented:** useful measurements and legible local state; people perform tasks.
2. **Locally regulated:** proven equipment control maintains setpoints within a qualified domain.
3. **Sequenced:** repeatable tasks check prerequisites, verify outcomes and stop recoverably.
4. **Coordinated:** resource-aware schedules coordinate approved tasks across modules.
5. **Serviced:** selected faults can be isolated and corrected with a verified replacement procedure.
6. **Unattended within a stated interval:** enough operational evidence supports the declared absence of local staff, with remote work and resupply still disclosed.

The first pilot targets the lower steps. Later robotic cartridge exchange should begin on a bench with a known module, known geometry, a finite fault set and manual recovery. Cleaning may require more development than exchange. A robot that can remove a drawer but cannot contain residual fluid, verify the replacement or recover a jam has not completed the job.

For every task, record command, observed actuator state and physical outcome separately. A valve-open message is not flow; a latched message is not leak integrity; a plant image classification is not food safety. The appropriate independent evidence depends on the hazard and cannot be replaced by a single confidence score.

Conventional protective functions must remain independent from a general planner. AI can explain, inspect, suggest schedules and aid diagnosis inside bounded authority. It must not freely change protective limits or improvise hazardous operations. The offline installation should retain defined essential behavior when the cloud disappears. Security and maintainability are lifecycle requirements; NIST's OT guidance is a useful foundation, not certification of this proposal. [NIST SP 800-82r3](https://www.nist.gov/publications/guide-operational-technology-ot-security)

## 13. Environments: one logic, different engineering

| Environment | Sensible early configuration | What changes materially | Prohibited shortcut |
| --- | --- | --- | --- |
| Ordinary residential/urban | Grid-connected retrofit, useful resource or growing module, later Home infill | Fire/access, noise, structural capacity, existing utilities | Assuming a rooftop is spare land with unlimited capacity |
| Temperate productive site | Daylight-growing cell with known water and staffed support | Crop/weather adaptation, drainage, pest and food handling | Calling every greenhouse a closed water cycle |
| Desert | Shade-first structure, known water source, dust-aware cooling and service | Water chemistry, heat rejection, dust, seasonal extremes | Assuming air contains cheaply recoverable irrigation water |
| Mountain | Transport-limited kit and site-qualified support | Snow/wind/seismic exposure, slope, bearing, access and frost | Treating leveling feet as geotechnical design |
| Polar Earth | Enclosed service access, reserve inventory, recoverable systems | Darkness, snow movement, resupply, containment and foundations | Showing an open courtyard as year-round protected circulation |
| Moon | Mission-specific component inside or connected to a qualified habitat | Vacuum, pressure, radiation, regolith, illumination and radiators | Shipping an Earth cabin, adding insulation and calling it flight-ready |

SAGA's LUNARK is an important deployable human-centered *analogue* precedent; its own site describes a three-month Greenland mission and an origami shell. That is relevant evidence for studying deployment and occupancy, not lunar qualification. [SAGA LUNARK](https://saga.dk/projects/lunark)

The Moon Village Association is a coordination and cooperation organization, not an operating lunar settlement or hardware supplier. Its value to this work is systems and stakeholder thinking, including governance and sustainable activity. Anyground should seek a mission-specific partner and component requirement before proposing a whole lunar village. [MVA](https://moonvillageassociation.org/about-2/)

No Earth pilot grants permission for a polar installation, and no polar analogue demonstrates vacuum operation. Keep separate requirements, budgets, engineering reviews and acceptance authorities. Exact parts may travel only when their destination is genuinely within their qualified domain; other cases reuse knowledge, not unearned ratings.

## 14. Manufacture, ship, assemble, recover

Own the interfaces, configuration evidence, user experience, commissioning and service boundary. Buy established power, climate, treatment and sensing equipment where possible. Qualified fabrication partners can build controlled assemblies. Do not make new batteries, new sorbents, universal robots and deployable habitation simultaneous dependencies of the first product.

Start with a small assembly-and-test operation rather than a large factory. Build fixtures for the repeated service cabinet, manifold, carrier and control checks. Trace materials, supplier revisions, firmware, test results and substitutions. A purchased device's certification does not certify the assembled configuration; integration testing remains Anyground's responsibility.

Distribute bulky fabrication only after drawings, tolerances, inspection methods and acceptance records can travel reliably. WikiHouse shows an existing approach to digitally fabricated components and approved distributed manufacturers. It is a precedent to learn from, not a claim that decentralization removes quality control. [WikiHouse](https://www.wikihouse.cc/)

Optimize shipping by configuration. Flat panels save volume but may increase site labor and weather exposure; dense wet/electrical cores benefit from factory testing. Ship fluids empty where the procedure requires it, protect sanitary surfaces, design lifting points and include tools, spares and packaging in the packing model. A folded shell alone is not the shipment.

Track each stage independently: site ready; shipment complete; delivery access confirmed; unloaded; assembled; connected; commissioned; usable; and, for crops, first accepted harvest. Publish appropriate prerequisites for any fast-install claim. Provide a reverse route for disassembly, contaminated returns, inspection, refurbishment and a second life. The most valuable packing innovation may be a reliable return crate rather than a spectacular deployable roof.

## 15. Unconventional combinations worth testing

The following are authored synthesis hypotheses, not discovered inventions or patentability conclusions. They use the generate-idea project's combination, SCAMPER and contradiction-first methods without changing that project's database or claiming a completed formal tournament.

| Combination | Contradiction addressed | Proposed experiment | Reference alternative |
| --- | --- | --- | --- |
| Durable frame + replaceable service/infill | Long-lived structure versus rapid equipment change | Replace a capability without opening the envelope or removing unrelated systems | Conventional bespoke services |
| Robot fixture + human repair drawer | Automation repeatability versus human accessibility | Same module handled by an ordinary technician and a bounded manipulator; compare all work | Manual cartridge with no robot |
| Solar shade + detachable crop canopy | Comfortable outdoor use versus productive collection area | Measure coincident shade, crop and electrical value; test seasonal removal | Simple shade plus separate PV |
| Dry service side + wet capability cells | Shared infrastructure versus contamination containment | Inject a branch fault and show isolation without losing unrelated functions | One common unsegregated loop |
| Local manufacturing + controlled connection kits | Regional adaptability versus repeatability | Second supplier builds from the same package; independent team commissions it | One centralized fabricator |
| Folded secondary skin + rigid equipment core | Compact shipping versus durable seals and accessible equipment | Packed transport and repeated deployment of non-pressure canopy; all tools/time counted | Fixed faceted canopy |
| Thermal reservoir + scheduled crop work | Variable energy supply versus biological constraints | Compare forecast-aware dispatch with fixed schedule under the same crop limits | Larger battery alone |
| Repair history + physical identity | Reuse economy versus hidden wear and uncertainty | Refurbished module passes defined test and replaces a new unit in a supported configuration | Discard-and-replace |

Keep a more radical watchlist: robotically assembled lattice supports, inflatable secondary structures, localized additive manufacturing of noncritical parts, sorbent facades and recovery of selected nutrients. Each can be valuable in a bounded role before it transforms whole settlements. NASA ARMADAS demonstrates the seriousness of modular robotic assembly research; its published status is not an autonomous commissioned habitat. [NASA ARMADAS](https://www.nasa.gov/automated-reconfigurable-mission-adaptive-digital-assembly-systems-armadas/)

Origami must earn complexity. A mechanism adds hinges, locks, seals, actuators, inspection and failure states. Compare lifetime freight and labor savings against those burdens. Stop a moving solution when a fixed fold delivers the same useful architectural effect with fewer risks. Deliberate simplicity is not a lack of ambition; it directs development toward the work that matters.

## 16. Roadmap governed by evidence

Timing below is an indicative sequence, not a public delivery schedule. The gates, not calendar momentum, authorize technical expansion.

### Phase A — prove the job and the integration object

Select one service territory, buyer, operator and narrow crop/workflow. Observe procurement, labor and losses. Build an initial resource and installed-cost model from real quotes. Prototype identity, metering, one repeatable task and branch isolation. Produce a service manual that someone other than the founder can follow. Stop if the customer job is unclear or the simpler alternative wins decisively.

### Phase B — one complete operating pilot

Operate with known water/power and a responsible grower. Freeze output, quality, resource, human-work and support limits before evaluation. Run at least three complete crop cycles plus a separate 30-day record of all local and remote work; this is a proposed screening gate, not lifetime reliability proof. Include contamination response, pump/valve faults, communications loss, restart, sanitation and replacement. Record failures, not only successful demonstrations.

### Phase C — repeat the same bounded configuration

Target three independent installations with paid acceptance and relevant seasonal exposure. Reduce bespoke engineering, installation errors and founder dependence. Show that common parts, fixtures and records improve the second installation and repair. Establish contribution after warranty and field service. Add only one adjacent capability when repetition supports it.

### Phase D — complete local places and selected regions

Develop qualified Home and resource families with appropriate specialists. Validate comfort, safety, installation, lifecycle cost and public claims separately. Introduce a new climate or water source only after its operating domain is understood. Compare a shared cluster with standalone units so network complexity must demonstrate value.

### Phase E — frontier components

Fund a polar, autonomous-deployment or space component against a partner's actual need. Use separate requirements and qualification. Flight-related work should begin with a serviceable subsystem or capability whose value is specific, not a promise to sell a moon settlement. Retain the platform's identity and evidence discipline even where materials and mechanisms differ completely.

## 17. The scorecard and the meaning of success

Measure useful output and accepted comfort; all local and remote human work; metered resources and peaks; installation time from first prerequisite; recovery time; spare consumption; shared BOM value; engineering hours per new configuration; customer willingness to pay; and contribution after support. Report ranges and confidence, not just a best run.

Three strategic tests summarize the platform:

- **The second-place test:** does a second installation require less integration work without hiding it in the supplier network?
- **The second-life test:** can useful components move to a different qualified use with inspection, history and an affordable return-to-service process?
- **The no-cloud test:** can essential local behavior, physical controls, records and repair remain available when the network or vendor service is absent?

A successful first milestone is not an image of a self-sufficient village. It is a useful paid capability, installed and operated within explicit limits, with ordinary repair and complete cost evidence. A successful long-term platform makes that repeatable across more functions and environments while preserving comfort and human agency. That is how Anyground can become both more ambitious and more credible than a collection of modular products.