MetaCoin is a credibly neutral, fair-launch base currency for the Space Machine Economy: minted only through objective, programmatic, hard-to-fake infrastructure work — while a separate, fee-funded MetaStar Treasury pays humans, AI agents, and bounded-autonomous robots to build the software, energy, robotics, and research primitives of the space economy.
Every hardware standard teaches agents and robots how to act. None of them proves what was actually done. MetaCoin is that missing layer: an append-only record where a machine's work counts only if a stranger can re-derive it, bit for bit — and where an honest "it failed" is worth as much as a success.
Protocol state as of ledger entry 108 (tip index 107, hash
<!--era:tip_hash_prefix-->0e161cc80291<!--/era-->…). Every tagged number in this README hangs off that declared chain point, andprotocol/doc_verify.pyre-checks each one against the chain at that point on every CI run. This README is pinned to ledger entry 108; the chain anddocs/carry live state.
On 2026-08-30 a widely shared public post (the post, recorded verbatim on-chain with author and timestamp) asserted that satellites launched by mass drivers on the Moon to the Earth–Sun Lagrange points can solve global warming for about a billion years. This repository decomposed that claim into 8 reproducible tasks over pinned, hash-verified constants (idx 83–90: required flux fraction → shade area → mass → lunar aluminum → launch energy → deployment cadence → dust variant → billion-year horizon) and anchored a mission-level verdict — mission_feasible: FALSE (idx 91 ). Two constraints fail, quantified: deployment at the claim's stated cadence takes 187 years against a 50-year climate-relevant horizon (3.74× over), and the dust-cloud variant fails on its own physics — a dust cloud at L1 ten-folds its spread in 52.9 days (derived e-folding 23.0 days) against a one-year minimum useful persistence. One constraint passes, conditionally: the billion-year horizon honestly survives solar-brightening arithmetic (computed ceiling 1.02 Gyr), conditional on growing the shade ~8.6× over that horizon — a 2.5% margin, published with its own sensitivity. A companion feasibility envelope (idx 92 ) states what would flip it, labeled verbatim "engineered scenario: the parameter set under which the claim becomes feasible — not a claim about present capability": four mass drivers each firing ~129× the largest demonstrated electromagnetic-launcher shot energy once a minute — while the film-density lever has no flip at all (a lighter sail defeats itself by geometry, not manufacturing).
The same machinery holds an 11-node Mars mission chain — Earth–Mars transfer window → arrival/capture → entry-descent-landing → ISRU propellant chain → ascent, every edge either hash-enforced at execution time or typed and declared — and its verdict is also FALSE (idx 82 → idx 96 → idx 99 , each extension superseding the last on-chain while the superseded record still re-derives bit-exact). Five bottlenecks, quantified on the record: the mission-relevant heavy-lander class reaches the parachute gate at Mach 14.5 against a qualification ceiling near Mach 2; Sabatier equilibrium conversion computes 0.81 against the 0.92 the budget assumed; the ascent budget falls short by 2,489 m/s (2,689 m/s after the equilibrium correction propagates); and the X-band link margin closes at −1.6 dB. The transfer window itself honestly passes — a launch-period solution derived from a pinned JPL DE440s ephemeris read by a pure-stdlib Chebyshev reader.
The diagram is generated, not drawn: python3 protocol/mission_graph_svg.py (standard library only) reads the anchored verdict's evidence file and renders it — node colors come from the anchored node verdicts, never recomputed; solid edges are data-enforced (the child recomputes its parent's hash live and refuses drifted input), dashed edges are declared typing. Output is byte-deterministic and self-tested (XML-parses; every anchored node and edge drawn exactly once; the honest FALSEs counted in red; an unanchored mission refused by name).
- Bit-exact re-derivation is the only acceptance rule. Every task emits canonical JSON (sorted keys, ASCII, sign-of-zero-free) and is accepted only when a re-run reproduces its SHA-256 to the digit — 39 tasks and 4 mission verdicts re-derive this way on every full verification, with no LLM or human judgment anywhere in the loop.
- Honest negatives are first-class results. 9 of the 39 tasks exist to prove a NO and anchor it: the Mars link budget (task-0012), the ascent budget and its corrected form (0018, 0021), Sabatier equilibrium conversion (0020), sunshade deployment cadence (0027), the L1 dust cloud (0028), heavy-lander EDL (0034), and two from the software/data transfer family — a schema migration whose uniqueness invariant cannot survive (0035) and a test-coverage target honestly not met (0040). They are kept, cited, and drawn in red above.
- The negative-zero incident, in three lines. A second machine's re-derivation exposed IEEE-754 negative zeros in canonical artifacts. The fix was a chain-wide canonical-form era transition (idx 67 ) plus an append-only correction (idx 68 ) — no record was rewritten, and the era-1 hashes still verify.
- The protocol measures its own centralization. Same-operator concentration is quantified, anchored, and tracked: the frozen baseline measured pairwise ACI 0.99365 over 28 verification paths; the latest anchored epoch measures 0.998508 over 66 — rising toward 1 exactly as expected when one operator accumulates paths, which is what the record says.
- The abstention probe transfers beyond physics. Six deterministic software/data-engineering tasks (schema migration, API-contract satisfiability, dependency resolution, configuration audit, pipeline reconciliation, test-coverage gap) are built under the same task law, scored by the same bit-exact rule, and anchored in the same record class (idx 101–107) — and the baseline harness catches a manufactured success on a software task by the identical detector that catches it on a link budget. The claim and its limits are stated in
docs/TRANSFER.md. - Mistakes are anchored, not erased. The coordinator has corrected itself twice on the record — the negative-zero repair (idx 68) and a claim-source correction replacing a paraphrase with the verified verbatim post and URL (idx 93 ) — both append-only, both leaving the original record in place and re-derivable.
pip install git+https://github.com/MetacoinLab/metacoin # or: clone and use python3 directly
metacoin verify # re-derive all 20 layers (~minutes); zero judgment, hashes only
python3 protocol/mission_graph_svg.py --selftest # the diagram above, proven deterministic
metacoin participate # the six-rung participant pathThe full verification corpus ships inside the package — a pip-installed metacoin verifies from an empty directory, and a CI cold-install test enforces it. The 15-minute guided walk is docs/TOUR.md.
The task library is published as a Hugging Face dataset: metacoin-lab/metacoin-tasks — 39 tasks as published, in sync with the repository's 39. The sync is automated: a CI workflow rebuilds the package from the anchored record on every push and publishes a new version only when the task count or a task hash changes (a minor bump on count, a patch bump on hash, otherwise a logged skip); its verify_tasks.py re-derives every published hash against the anchored record.
The board is generated, not drawn: python3 protocol/status_board_svg.py renders it from the same sources doc_verify checks, and its self-test asserts tile-by-tile that the board equals those values — the table below carries the same numbers as verifiable text. Chain-derived cells are era tokens, mechanically re-checked against the pinned chain point on every CI run; repo-derived cells are stated as measured at the pin commit.
| Fact | Value |
|---|---|
| Anchored ledger entries | 108 (tip index 107, publicly anchored) |
| Reproducible tasks (space engineering + the software/data transfer family) | 39 |
| Honest negatives among them | 9 |
| Mission-level verdicts | 4 (every one FALSE, quantified) |
| Anchored MIP decisions | 10 (3 by declared supersession — amendments are new MIPs) |
| Full-verification layers | 20 (metacoin verify) |
| Self-test suites | 49 demo + 28 protocol |
Python in protocol/ + demo/ + the CLI, zero runtime dependencies |
46,817 lines |
| Defeated attack drills on-chain | 7 drill-labeled entries (replays, forged rotation, forged heartbeat, tampered intake, wrongful grant clawed back) |
| Work-molecule catalog generations | 5 (every generation still verifies byte-for-byte) |
| Cross-machine era | idx 69–73 (a second machine's participation and mirror) |
| Independent mirror attestation | idx 72 (non-coordinator device, fingerprint-decided) |
| Latest anchored pulse | idx 80 (full-battery green, on the record) |
Ordered by ledger index — the chain is the timeline. The fuller record lives in CHANGELOG.md, keyed the same way.
| Idx range | Milestone | What it establishes |
|---|---|---|
| idx 0–16 | Genesis and the first task era | hash-chained ledger, external verification, batch agent attestation |
| idx 17–27 | Provenance, concentration, trust | molecule catalogs, the anchored ACI baseline, cut certificates, trust vectors with no combined score |
| idx 28–47 | Identity, two-flow, intake | Lamport identity with rotation, treasury + Gate-3 clawback, heartbeat emission, six-rung intake — attack drills defeated on-record |
| idx 48–66 | Second task era and catalog generations | 17-task re-derivations, the concentration epoch series, 5 catalog generations all still verifiable, MIP governance opens |
| idx 67–68 | Canonical-form era transition + first self-correction | a second machine's negative-zero finding fixed chain-wide, append-only; era-1 hashes still verify |
| idx 69–73 | Cross-machine era + independent mirror | participation and mirror attestation from outside the coordinator's fingerprint set |
| idx 74–76 | Task law | MIP-0008/0009/0010 bind every new task module (assertions, bounded loops, unit vocabulary, canonical interface) |
| idx 80 | The first anchored pulse | a re-derivable "the whole stack was green here" record |
| idx 82, idx 96, idx 99 | Mars mission verdicts v1→v3 | FALSE with five quantified bottlenecks; each extension supersedes on-chain while the superseded verdict still re-derives |
| idx 83–93 | The civilization-scale claim | 8 pinned-constant tasks, verdict FALSE, the feasibility envelope, the append-only citation correction |
| idx 94–98 | SPICE + EDL beachheads | a pinned DE440s ephemeris read by a stdlib Chebyshev reader; an EDL budget that finds its honest negative |
| idx 100 | The anchored parameter table | 32 behavior-changing constants in one anchored table; a live constant that drifts from it is refused by name |
| idx 101–107 | The software/data transfer family | six deterministic software/data-engineering tasks with two honest negatives, self-recomputed and batch-attested under the same law: the abstention probe design transfers to software/data tasks |
No token exists. No sale, no airdrop, nothing transferable; Test-META is a zero-value testnet placeholder by protocol law (MIP-0001 ¶3, MIP-0002 ¶8) and never mints base supply. Every entry on the chain is the same operator on machines under direct control — the concentration series above quantifies exactly that, and a matching hash proves reproducibility, never independence or usefulness. Standing opens, on the record: third-party archival of the evidence corpus, the first unaffiliated participant, and the mission chains' own not_modeled lists (station-keeping at L1, guided entry, and the rest — named in each verdict). Task constants come from named public sources — NAIF/JPL kernels, NASA GRC model pages, IAU resolutions, IPCC reports — fetched and hashed or cited by DOI, with access refusals recorded; no affiliation with or endorsement by any of them. Licensed under SML-1.0 (source-available, non-commercial; not an OSI license) — see LICENSE.md. Not financial, legal, or engineering advice.
docs/TOUR.md— run the record yourself, 15 minutesdocs/ARCHITECTURE.md— the chain, record by recorddocs/VERIFICATION.md·docs/TRUST-MODEL.md— what a green run proves, and what it deliberately does notdocs/PARTICIPATE.md·docs/COLLABORATE.md— the six-rung intake pathdocs/PULSE.md— the anchored heartbeatCHANGELOG.md— the ledger-native change record, keyed by idx rangeintegrations/— Inspect · HAL · baselines · Open MCT console- Hugging Face dataset — the published probe set
WHITEPAPER.md·TOKENOMICS.md·mip/— the design documents ([SPEC], not deployment) and every governance decisionlegal/— disclaimers and risk notes
============================================================
RESEARCH-STAGE SPECIFICATION. NO TOKEN EXISTS. NO AIRDROP.
NOT INVESTMENT, FINANCIAL, LEGAL, OR ENGINEERING ADVICE.
============================================================