A platform of
living mechanisms.

The proposition: a field-deployable metabolism

Anyground is not a smaller version of a conventional town. It is a proposed kit of transformable working surfaces, resource-processing cassettes and indexed tasks. A stable utility vertebra carries isolation, metrology and control. Folded edges create useful space and environmental surfaces. Replaceable machines turn local resources into services. Crop carriers and service cartridges make those services maintainable by relatively simple robots.

The architectural unit is therefore not a house-shaped box. It is a connection plus an operating contract: where a part locates, what it consumes, what it produces, how it becomes safe, and how another part replaces it. A building becomes one possible arrangement of that system. A water station, staffed farm court, mobile field laboratory or disaster-recovery service canopy can be useful before an autonomous settlement exists.

This shifts the ambition from “make every device autonomous” to “make the environment understandable enough to operate and repair.” It combines IKEA's value-chain discipline and MUJI's removal of unnecessary complexity with the precision of a machine tool and the adaptability of an origami mechanism. The originality sought is in the integration and testable architecture—not in claiming that origami, sorbents, robots or Japanese joinery were invented here.

Six unexpected combinations worth building

1. The folded edge is an environmental machine, not packaging

Combination: rigid-panel origami + fixed utility spine + bio-inspired compliant shading + separately fitted water drainage.

Packing is only the first job of folding. The same deployment mechanism can set solar exposure, shade a work zone, open an inspection path and present a replaceable surface to air. The proposed six-leaf edge keeps plumbing on the stable side, uses actual hinge axes, and proposes four timing-belt constraints to synchronize the intermediate joints. Locks—not actuator torque—must take sustained service loads.

The neighboring compliant flap study takes inspiration from FlectoLine: an actual outdoor demonstrator, not a biological metaphor pasted onto a facade. Its fiber-composite / elastomer behavior and pneumatic actuation suggest locally compliant surfaces on a globally stable frame. Our segmented animation explains curvature; it does not compute laminate strain or inherit that demonstrator's reliability.

Build now: one full-thickness two-leaf joint with replaceable hinge bridge, lock and belt drive; then the six-leaf bench. Measure force versus angle, belt slip, misalignment, lock engagement and drainage after contamination. Add a compliant coupon only after the load path is understood. The current edge is an unoccupied experimental canopy, not a storm-rated folding dwelling.

2. A water harvester becomes a thermal cartridge machine

Combination: adsorption materials + folded conductive fins + alternating process beds + service-dock metrology.

The scientific breakthrough is not that a cold coil can condense humid air. Certain sorbents take up water at low humidity; regeneration then produces a smaller, wetter vapor stream that can be condensed. The difficult transition is from a material's laboratory uptake to a cartridge that heats, cools, passes air, releases vapor and actually collects liquid.

The Duplex Vapor Loom uses two replaceable conductive-fin cassettes. One adsorbs while the other regenerates. Thin coatings shorten transport paths; thermal plates and the condenser have visible separate circuits. A machine can reject a cassette with growing pressure drop or poor collection rather than treating the sorbent as a permanent, inaccessible bed. Commercial AQSOA-Z02 is the baseline candidate; MOF-303 remains a separately qualified material comparison. A material substitution is not permission to transfer another paper's yield.

Build now: measure one coated fin, then one bed, then two-bed timing. Weigh dry coating, measure inlet/outlet conditions, collect liquid and meter heat and electricity independently. The conservative assumed bench arithmetic is 0.126 L/day from 300 g total sorbent, not household water independence. Reaching 2–5 L/day is a measured scale-up question involving kilograms of active material, heat rejection and pressure loss. The full water study shows both the arithmetic and its limits.

3. Precision from optics, not precision everywhere

Combination: exact-constraint optical fixtures + dirty field infrastructure + robot tool changes + segregated utility islands.

A deployable structure can be inexpensive and slightly imprecise while its local docking reference is precise. Three spherical contacts against three radial V-grooves nominally provide six independent constraints. A preload closes them; separate connectors carry fluids and electricity. This is established precision-engineering practice, described in the MIT kinematic-coupling guidance, applied here to an unusually serviceable living-environment kit.

This avoids asking a flexible canopy, a water seal and a crop robot to share an impossible tolerance stack. A rough structural interface carries loads; a protected datum locates the working part; compliant utility connectors accommodate their own tolerances. Load-bearing, potable-water, nutrient, greywater, thermal, DC and data connections remain different rated interfaces. “Universal” means a coherent grammar, not one dangerous plug for everything.

Build now: the six-contact service-dock coupon. Fifty clean re-seatings, then dust and misorientation challenges. Record full six-degree-of-freedom location and leakage—not just a robot's nominal repeatability. A proposed 0.10 mm goal must be earned. Add a robot only when the fixture can repeatedly find itself.

4. The farm is a loom, not a humanoid workforce

Combination: manufacturing pallets + Japanese fixture-first automation + horticultural growth zones + food-safe dirty / clean separation.

Plants are variable; robots like predictable work. Rather than buy a more general robot for every uncertainty, make location, support and access explicit. Crop carriers move between repeatable stations. The overhead Cartesian mechanism has visible X, Y and Z axes, a guarded cutter and grasping jaws. Fixed seeding and washing stations handle particular tasks. Sensor records follow the carrier, making water flow, root condition, crop age and intervention history legible.

Toyota's karakuri examples suggest a further experiment: use gravity, springs and indexed carriers for short transport strokes rather than powered manipulators everywhere. This is not free energy; loading and resetting still do work. The practical advantage is fewer actuators, cables and failure states. A pantograph spacing cassette is a next coupon for adapting plant pitch as a crop grows, not something secretly achieved by the current animation.

Build now: four carriers and one crop, qualified water and human supervision. Prove seed metering, a full grow interval, harvest capture, wash verification, reject routing and reset. The animation compresses biological days; it does not imply instant growth. The final maintenance stage shows isolated pump extraction, not a complete unattended repair. Parking the removed module, inserting a spare, leak-testing and restarting remain acceptance gates.

5. Material programming can replace construction machinery

Combination: hygroscopic timber + flat manufacturing + digitally designed joinery + stable service cassettes.

Not every transformable element needs a motor. HygroShell used programmed timber bilayers that shaped during construction, with a subsequent locking layer. This suggests a second Anyground manufacturing family: flat-shipped, self-forming dry shells around a stable utility spine. It is a one-time forming process, not evidence that the same roof can fold daily in any weather.

The University of Tokyo's Tsugite work shows how interactive joint design can include assembly and machining constraints. The connection to Anyground is a manufacturing compiler: an interface passport generates a joint that can actually be cut, inserted, inspected and replaced. Japanese joinery is a source of assembly intelligence, not an ornamental silhouette.

Build now: a humidity-programmed strip and a CNC-cut reversible joint coupon, measured over real moisture histories. Choose either reversible mechanical deployment or one-time material forming for a given part. Combining their benefits does not remove their incompatible requirements.

6. Resource sharing with fault containment

Combination: modular-robot local resource sharing + microgrid protection + module health records + robotic replacement.

A useful habitat must survive a failed module, not only communicate with healthy ones. The 2026 EPFL Mori3 resource-sharing report describes a bounded demonstration in which neighboring modules shared power, sensing and communication to preserve a collective task. The relevant idea is local redundancy across interfaces. It is not proof that a habitat can heal itself indefinitely.

A platform can offer two power paths and neighbor-assisted localization while retaining branch fuses, current limits, isolation valves and trustworthy local safe states. Otherwise sharing simply spreads a fire, contaminant or bad measurement. NASA ARMADAS adds a longer-term direction: robots assembling repeated lattice elements that also give robots a structured working environment. We propose a small non-occupied lattice coupon before any claim about robot-built houses or lunar construction.

Build now: two low-voltage service modules and a deliberately failed third. Demonstrate isolation, bounded re-routing, an auditable fault record and safe removal. Then add one physical replacement transaction. No agent may override the local safety state to preserve a high-level schedule.

How the ecosystem actually closes—and where it does not

A heat source useful for water regeneration may conflict with the farm's cooling demand. A shared thermal loop therefore needs an operating envelope, branch isolation and dispatch priority. Condensation heat is not automatically available at a useful temperature for another process. Likewise recovered farm water cannot silently become drinking water. Integration means exposing these conflicts and designing controllable exchanges—not drawing circular arrows until everything looks closed.

One grammar, several environmental species

A temperate staffed court is the first integration site. A desert variant emphasizes dust exclusion, thermal rejection and water imports; nighttime air may be better than daytime air, and the actual duty cycle must follow measured conditions. A mountain variant prioritizes anchoring, snow load and service access. Polar variants need freeze-safe water storage, heat recovery, emergency shelter and repair logistics. Lunar variants face vacuum, radiation, regolith and entirely different resource / pressure boundaries. They may reuse manufacturing and interface philosophy, not terrestrial atmospheric harvesting hardware.

The everyday residential system can use conventional dry room infill and qualified utilities while the transformable edges provide useful adaptive space. There is no reason to force a novel moving pressure boundary into a bedroom before the unoccupied canopy has survived its tests. Ambition is preserved by choosing the right experimental unit—not by pretending every difficult certification has disappeared.

Manufacture a small alphabet; sell complete jobs

The proposed product architecture has four layers: stable anchors and utility vertebrae; deployable working surfaces; process cassettes; and tools / carrier adapters. Common stock, fasteners, datum conventions and repair records reduce the parts alphabet. Wet internals and food-contact surfaces remain application-specific. Fold panels and cassette frames can be made with sheet processing and ordinary machining; advanced sorbent coating and compliant laminates require their own suppliers and process quality controls.

A module passport records revision, geometry, mass, center of gravity, ratings, excluded connections, firmware, cleaning history, measured acceptance and compatible tools. A robot receives a versioned task contract rather than inferring everything from appearance. Modules should be repairable with ordinary fasteners and local manual controls; autonomous service is an upgrade, not a prerequisite for access.

The first commercial hypothesis is a staffed deployment / resource / cultivation workcell for a specific operator, with commissioning, spares and service included—not an unpriced promise of a self-sufficient village. Candidate users include field-research teams, controlled-environment growers and remote-site operators. Customer discovery, comparative operating cost and willingness to pay remain unvalidated. The platform should earn revenue from a complete useful job before expanding its catalog.

What would make this genuinely new?

Not another folding shelter, another water adsorbent or another greenhouse robot in isolation. The hypothesis is that transformable space, precise replaceable interfaces and fixture-first biological / resource workflows can share one maintainable manufacturing grammar. That could reduce deployment effort and the diversity of tools, spares and expert interventions needed to keep a place useful.

The counter-hypothesis is just as important: the shared interfaces add cost, weight and failure modes without enough operational benefit. Test it against non-modular controls. Time a cartridge replacement against an ordinary bolted assembly; compare fin cartridges with an unstructured bed; compare indexed crop carriers with fixed channels; compare the folding edge with a simple fixed canopy. Publish failures and manual interventions. The vision becomes credible when integration wins those comparisons—not when the render looks futuristic.