A testbed for the question underneath Technofeudalism / Another Now: can a post-capitalist economy hold together without a gun?
The wager, derived over a long design conversation:
- A system that must be maintained by force — a wall to stop people leaving — is a system of force. That's communism's flaw, and it's disqualifying.
- So "does the solution work" is the same test as "does it hold with the door open": run the economy with a genuinely competitive, neutral exit always reachable, and see whether free agents elect to stay.
- Retention-under-open-exit is therefore the loss function — and, unusually, it is normatively identical to the goal. It's also a sufficient statistic: an inoperable, unstable, or coercive economy all bleed agents to the exit.
- The exit lives outside the optimizer's reach, always. The moment the exit is a tunable parameter, you're optimizing a wall with the lights off.
We test the destination first (does it hold when agents are dropped in), independently of the transition (how you'd get there) — because a destination that collapses makes transition study pointless, and a destination that holds under an open door has already shown adoption can be voluntary.
Three layers, with the exit as a frozen boundary between them:
- World (this repo, Stage 0): a deterministic, LLM-free economy. A
Ledgerof typed assets with one settlement primitive (transfer);Sovereigntyimplementations (a VaroufakianCommonswith per-capita issuance + demurrage + demurrage-funded dividend; a frozen neutralMarket); aContractlayer (Stage 0:SpotTrade; later: combinators for credit, escrow, futures — everything reducing to ledger transfers); a minimal goods economy so money has a job. - Agents (Stage 1+): scripted policies now; LLM "free-will" policies later,
behind the same
ordersinterface. Tiered — a cheap crowd, strong adversarial probes. - Optimizer (Stage 4): black-box search over commons parameters to minimize the regularized retention loss, with population holdout. The exit is never in the search space.
| Stage | What | Gate |
|---|---|---|
| 0 | general abstractions, one commons vs frozen market, scripted agents | mechanics + circulation + exit neutrality |
| 1 | LLM probes at a hand-tuned config | operability / stability / election, descriptively |
| 1.5 | many sovereignties (the design tournament) | — |
| 2 | adversarial red-team (Sybil, dividend-farm, shadow currency, capture) | no instantly-fatal exploit |
| 2.5 | Kosmos inter-sovereignty clearing (symmetric penalty + exclusion) | federation holds vs. first defector |
| 3 | open-ended agent-minted good-tokens | — |
| 3.5 | contract combinators (diverse trade) | — |
| 4 | optimization + report | params clear the market baseline on holdout — or none do |
cd disencloser
python run.py
Prints a two-economy demo, the redistribution check, and the three Stage-0 gates. No dependencies beyond the Python standard library.