Deep-Time Preservation of Human Knowledge — Research Round 1

Memory Commons reference library · 5 July 2026 · adversarially verified deep-research run (106 agents, 24 sources, 118 claims extracted, 25 verified: 22 confirmed, 3 refuted). Raw verified findings: 2026-07-05-deep-time-preservation-round1-findings.json.

1. What has already been attempted

EchoStar XVI “The Last Pictures” (2012) — the closest precedent to a Memory Commons satellite artifact: a passive encoded object attached to a commercial GEO satellite. Two gold-plated aluminum jackets hold a silicon disc with 100 nano-etched photographs. The cover is a self-describing primer: time defined via the neutral-hydrogen hyperfine transition (1,420,405,751.768 Hz, Pioneer-plaque convention), epoch encoded via 13 pulsars to ~0.02-day accuracy (Weisberg & Paglen, Astronomical Journal — arxiv.org/pdf/1208.4637).

Rosetta Disk (Long Now, 2008) — ~13,000 microetched analog pages covering 1,500+ languages on nickel, readable with optical magnification alone; no format or platform dependency (rosettaproject.org/disk/concept). The decodability gold standard.

GitHub Arctic Code Vault (2020) — 21 TB on 186 reels of piqlFilm in a Svalbard mine. Selection was rule-based (any repo with recent commits; ≥1 star + activity; ≥250 stars), not curated (archiveprogram.github.com/arctic-vault).

Arch Mission Foundation — nickel Nanofiche libraries: crashed with Beresheet on the Moon (2019), 5D-quartz Asimov trilogy on the Tesla Roadster escape trajectory. Vendor projects billions of years in space, ≥10,000 years on Earth — unaudited foundation projections backed only by short accelerated-aging tests (archmission.org/nanofiche).

“Message in a Bottle” (Jiang et al. 2023, peer-reviewed) — successor proposal to the Voyager Golden Record. Documents the Record’s failure modes (ethnocentric content, 1970s encoding) and proposes a two-tier design: analog scroll readable without technology + digital layer whose access procedure is symbolically taught, each layer the key to the next (AGU, DOI 10.1029/2023EA003042).

Named but not verified this round (research round 2 covers them): Voyager/Pioneer specifics, LAGEOS, KEO, Memory of Mankind, Sanctuary on the Moon, Celestis/LifeShip, Project Silica.

2. If Earth disappears: what remains, how long, found by whom

Premise correction: a catastrophe does not “propel satellites to safety” — they are already in space, and no Earth catastrophe meaningfully changes their orbits. Whatever is in a stable orbit simply stays there. Design question: which orbit is already safe.

Verified persistence ladder, worst → best:

• LEO — worst and legally capped. Uncontrolled decay in years–decades; FCC 5-year rule (adopted 29 Sep 2022) caps US-licensed LEO satellites at mission-life + 5 years (FCC 22-74; 47 CFR 25.283(e)). Repeal petition pending (docket GN 25-133) as of mid-2026.
• MEO — disposal orbits unstable via lunisolar resonances; re-entry within ~40–200 years.
• GEO / graveyard — no appreciable drag, persists essentially in perpetuity; ceiling set by solar evolution (~7.6 Gyr to red giant reaching Earth’s orbit).
• Lunar surface — size-dependent abrasion; smaller is better: ~0.2 mm/Myr for cm-scale vs ~20 mm/Myr for 10 m blocks (Icarus 2022). A thin low-profile plate survives tens of Myr exposed; longer buried.
• Escape trajectories — longest survival, least discoverable.

Discovery probability: no defensible number exists; for unknown intelligence it is effectively zero per artifact (small, dark, passive). The probability story changes completely if the audience is our own successors — human or AI — who inherit the catalog of artifact locations. This reframing is the project’s strongest card: the physical artifacts are the disaster-recovery tier of a live, digital, identity-anchored commons whose index records where every copy sits.

3. Storage media — verified state, hype stripped

• Nickel Nanofiche — analog, ~2,000 pages/cm² at 150 dpi, no decoding standard needed, commercial mastering today. Longevity figures are vendor projections. Best Tier-1 medium.
• 5D fused silica (Southampton/Kazansky) — 360 TB/disc and “unlimited lifetime” are projections; the 2013 demo stored 3 bits/voxel (eprints.soton.ac.uk/364916). Later work ~4 bits/voxel. Durable but needs a polarization microscope — Tier-2 medium.
• Tungsten in silicon nitride — real physics behind gigayear claims: ~1.8 eV Arrhenius barrier suffices for 1 Gyr; samples survived accelerated aging implying >1 Myr error-free (arXiv:1310.2961 — preprint, extrapolated). Nested-QR self-similar encoding worth adopting.
• DNA — verified as promising in density; economics and fragility rule it out for space; optional for cold vault redundancy.

Recommended encoding methodology (MIAB two-tier + Weisberg–Paglen primer):

• Tier 0 (etched, naked-eye/low magnification): hydrogen-line time unit, pulsar epoch map, binary tutorial, map of other artifacts.
• Tier 1 (Nanofiche analog): foundational set — math, physics constants, genetic code, languages Rosetta-style, images of human life, music as notation + waveform diagrams, chemistry of the senses (odorant molecules), a charter of who we were and what we valued.
• Tier 2 (silica/tungsten digital): full curated corpus; decoding procedure taught pictorially in Tier 1.

4. MC-1 artifact concept (working spec)

~100 × 100 × 3 mm, <300 g, fully passive. Stack: etched primer face (tungsten/Si₃N₄ or engraved nickel) + Nanofiche analog library + one fused-silica digital wafer, in a gold-plated or ceramic jacket, deliberately regular/engraved so a finder recognizes intent. Bolt-on to exterior structure or ballast mass; no interface to the host bus — which is what makes OEM adoption plausible. Placement mix: GEO hosted payloads (workhorse), lunar landers (discoverability), escape trajectories (deep time), terrestrial vaults (reachability). LEO excluded.

5. Funding & scale-up (UNVERIFIED — no claims survived adversarial review round 1; analysis only)

Proven pattern: Celestis/LifeShip-style paid personal inclusion (~$100 DNA/name to Moon → thousands for memorials). Loop: each plate reserves Tier-2 capacity for paid personal messages, funding the curated commons payload that rides free. Complements: per-mission corporate sponsorship (GitHub/Piql model), institutional civilization-backup customers, grants. Rideshare ~$6–9k/kg LEO (unverified 2026 pricing); first GEO artifact plausibly $50–150k all-in. Scale-up = publish the MC-1 spec + qualification pack openly so any OEM can carry one; make it a badge.

6. Authenticity & governance (design proposal; round-1 claims did not survive verification)

Every corpus item carries a C2PA-style provenance manifest signed by a contributor DID (AltersID registry). Every corpus release is a Merkle tree whose root hash is physically etched on every plate — any recovered artifact authenticates any surviving digital copy; any two artifacts verify each other. Crypto only needs to survive between plate generations; the etching is the ultimate trust anchor. Anti-capture: public transparency log of corpus versions; two-track selection (rule-based baseline + curated layer via rotating multi-regional panels with open challenge); no-single-donor-dominance; multi-steward release keys. Voyager’s verified lesson (via MIAB): elite curation is the reputational failure mode.

7. Critical feasibility conclusion

Technically feasible with flown precedent; nothing requires invention — integration, qualification, and tens-of-thousands-scale money for a pilot. Three honest corrections: (1) catastrophe-ejection premise replaced by orbit selection; (2) all “billion-year” media figures are projections — say “millions credibly, billions plausibly”; (3) alien discovery probability cannot anchor the pitch. The strong, novel position: an archive whose primary intended reader is future machine intelligence, coupled to a live identity-anchored digital commons — fixes discovery (successors inherit the catalog), fixes decodability (the commons carries the decoding stack), and gives physical plates a coherent role as the write-once trust anchor. Hardest problems are non-technical: content-selection legitimacy, decades-scale institutional survival (KEO), and keeping funding from distorting the corpus.

Refuted claims (0–3 votes — do not use)

• 5D silica 2013 paper being the first digital 5D recording.
• Nanofiche 1-micron character size implying purely optical recovery.
• Micrometeoroid bombardment as the dominant lunar fillet-forming erosion mechanism.

Open questions carried to round 2

Complete survey of remaining projects; verified rideshare/hosted-payload and lunar delivery costs; verified funding models (Celestis pricing, Long Now/Internet Archive/Software Heritage budgets); C2PA adoption + long-term signature standards (RFC 3161, ETSI LTV, NIST post-quantum); Ostrom-style governance precedents; graveyard-orbit stability studies; lunar lava tubes; materials cost/GB matrix (Cerabyte, Arnano sapphire, Project Silica, piqlFilm).