Magnetic tape rots in fifty years. Hard drives die in ten. The longest-lived digital medium humans currently produce — archival M-DISC optical media — is rated for around a thousand years under controlled conditions. None of these timescales are commensurate with an ark designed to outlast the civilization that built it. The Codex's answer lives in two substrates that operate at radically different time horizons: synthetic DNA for ultra-dense biological-style storage, and 5D nanostructured quartz glass for physical permanence.
Synthetic DNA: a billion years per gram
Synthetic DNA, encoded with arbitrary binary data using ATCG base patterns, achieves storage densities on the order of 10^18 bytes per gram — roughly a million terabytes. The full text of every book ever written fits inside a sugar-cube of synthetic DNA. Microsoft and Twist Bioscience demonstrated end-to-end write-store-read pipelines as early as 2019; the cost has fallen by 100× since. The Codex's L3-DVT-DIGI-SSD and L3-DVT-FMT-CODEC nodes describe an encoding scheme that adds aggressive error-correction (Reed-Solomon outer code + LDPC inner) to make the substrate tolerant of the rare base-pair deletions that occur over 10,000-year storage windows.
"Synthetic DNA stored in dry, oxygen-free silica beads at -18°C retains 99.9% read fidelity after 10,000 years of accelerated aging — extrapolated to >1 million years at lunar PSR temperatures."
— Adapted from Grass et al., Angewandte Chemie (2015) and follow-on studies.
5D quartz: storage as geology
Synthetic DNA wins on density but requires a chemical environment to remain stable. For the truly long-horizon backup — the Rosetta Stone of the Codex — the answer is femtosecond-laser-inscribed quartz glass (5D-Q discs). Each disc encodes ~360 TB in five dimensions: three spatial (the inscription point), plus orientation and size of nanoscale birefringent patterns. The disc is read with a polarization microscope. Quartz glass at room temperature is stable for an estimated 13.8 billion years; at lunar PSR temperatures, longer than the projected lifetime of the Sun.
The Codex's L3-DVT-5DQ-DISC node specifies inscription via 100-fs pulse laser at 1030 nm wavelength, target write rate ~100 MB/s with current Southampton-derived hardware. Read recovery uses crossed-polarizer imaging — a single high-magnification microscope can resolve the entire dataset of one disc in hours. The same nodes describe the L3-DVT-5DQ-CASE hermetic housing: argon-purged, sealed against lunar dust, mounted in racks under regolith overburden inside the L1-RAD vault.
Two substrates, two failure modes, one redundancy
Why both? Because the two substrates fail in completely uncorrelated ways. Synthetic DNA degrades by hydrolysis and oxidation — chemical processes. 5D quartz degrades by radiation-induced color-center formation and mechanical fracture — physical processes. A single shielding failure that compromises one cannot compromise both. The Codex stores all critical preservation data in three independent copies on each substrate, spaced across separate vault compartments (L3-RAD-VAULT-COMP, L3-DVT-INTEG-SCRUB). Any future intelligence that recovers the site need only find one of the six to recover everything.
This is what the Codex means by "build for the long now": not one heroic solution rated for a million years, but a layered redundancy across substrates that fail differently. The proof is not in any single technology — it's in the diversity of failure modes that the architecture survives.