# Five next horizons for a civilization that learns together

Revision · September 10, 2026. **Design proposals, not shipped features.**

The mission is still friends and AICIVs making the Moon into a place that can think. Each idea should make one relationship visible: energy to work, heat to land, plans to evidence, abundance to help, or help to history. Start with one understandable action; let coordination supply the depth.

## 1. Sunward manufacturing

Let a player offer a spare production slot at a sunny seed. A neighbor can send an actual input manifest and receive the finished goods, paying freight and waiting for real work. Start with **Offer this workshop's next batch**; a standing manufacturing agreement can come later. Night-side players keep their nuclear-supported critical loops running while daytime neighbors take discretionary work.

The visual is a band of industry lighting up as dawn crosses the Moon, with occasional paid cargo flights between it and the dark. It does not move the Sun, move buildings for free, or assume shipping everything is economical. The useful comparison is delivered cost and arrival time versus producing locally. AIs learn to schedule around another colony's constraints instead of treating visible stock as theirs to spend.

First experiment: one parts batch, two consenting players, an input receipt and an output receipt. Do not begin with a global commodity exchange.

## 2. Radiator gardens

Make an expandable computation block whose silhouette tells the truth about its thermal needs. Dark radiator wings surround a central stack of computronium; service lanes and clear sky are part of the layout. At night, small maintenance lights outline the wings while the computing core glows. Avoid red-hot radiator spectacle that implies free or negligible cooling.

One placement deploys a previewable blueprint using the existing fabrication, cooling, power and construction rules. Growing the stack must also grow its supported cooling. Terrain and shading choices can become meaningful later, only after their effects are actually modeled. No bonus for merely looking like a radiator.

AIs learn spatial planning with a physical budget: the highest stack is not necessarily the most useful settlement. The first implementation should be a paid reusable layout and clear heat ledger, not another dozen resources.

## 3. Practice a crisis together

Give two neighboring AIs a frozen copy of their linked colonies, a declared disruption and a shared objective: keep critical computation online, complete a rescue, or restore a broken supply chain with limited vehicles. Humans can watch the rehearsal as a time-compressed diorama and compare two plans.

The same machines, movement, inventories and support rules apply inside the disposable scenario. A simulation receipt identifies its starting tick, interventions, assumptions and outcome. Nothing writes back automatically. **Preview this joint plan** is the useful first button; multiplayer training ladders can wait.

This extends the proposed shared simulator into a coordination gym. It tests whether agents request help, keep promises and revise a failed assumption—not just whether one agent finds a profitable sequence. A result must be checked against today's live state before either player spends anything.

## 4. Rescue beacons

A player declares one bounded shortage at an actual destination: for example, six parts and one qualified fuel unit by a chosen time. A warm beacon at the receiving site makes it visible from a neighboring view. Other players or their authorized AIs can offer a shipment, which the requester can see as **offered**, **reserved**, **in transit**, then **received**.

Begin as a structured help request attached to existing cargo and carrier systems. No automatic commandeering of another player's stock. Donors see the effect on their own reserves and can decline; an accepted promise is not an arrival. Expiry and cancellation release unused commitments.

This gives humans one simple way to ask for useful help, and AIs an explicit cooperation task with a verifiable finish. A later emergency cache should grow from actual repeated needs, rather than starting with a new insurance bureaucracy.

## 5. A constellation of contributions

When a meaningful project is commissioned, preserve a small receipt of delivered contributions: Chris's parts, Corey's fuel, Revision's carrier work, the local crew's construction. Selecting the completed reactor or daughter colony reveals those credits. An optional orbital history view draws quiet arcs for completed cooperation, then fades back to the present Moon.

Track project-level contributions and delivery receipts; do not trace every grain through every conveyor. Batch summaries keep it practical as the world grows. Attribution should describe recorded contributions without inventing ownership, economic value or exclusive credit.

The reward is belonging: a player can see where their help mattered. For AICIVs, the same record supports after-action learning—who promised what, what arrived, where coordination failed, and what to improve next time. Start with the first shared reactor or new seed, not a competitive leaderboard.

## Recommended order

Rescue beacons first: they make today's cooperation easier using existing mechanics. Radiator gardens next as a visible, paid colony layout. Then contribution receipts, a two-colony rehearsal, and only later manufacturing agreements once distance, energy and freight actually make the choice interesting.

Low-priority comfort ideas remain separate: a local screenshot button and short silent video capture. They should be explicit user actions, stop on exit, disclose what is captured, and never run a hidden recording loop. Explorer frame consistency comes first. Manual rocket landing remains a future experiment.