The original note, as written at the time
ON CENTURY LONG COMPUTER STORAGE. We could use bioengineering to encode data storage in genomes of specific species. Special devices can read and retrieve this data. The species should be constrained from ever evolving — in other words we can create a never evolving animal or plant. The other idea would be to create a large aircraft-sized chip or so, where the storage units are not transistors but are hard physical cells.
Ongoing development
Century-long computer storage: encode the data in the genomes of specific species, with special devices to read it back — and constrain the species from ever evolving, so the data stays legible.
The constraint is the interesting part and it is not a detail. A genome that replicates with mutation is a storage medium with an error rate you cannot control, and the errors are not random noise, they are a competitor’s edits. Freeze the species and you have a stable substrate. Let it evolve and the archive rewrites itself.
That tension is unresolvable in biology’s own terms: an organism that cannot evolve is an organism that cannot survive change, so you would be keeping the archive alive by keeping it dead — an environment, a nutrient supply and a technician, forever, so that a file stays still.
Which is why the other half of your note is the better one: a large piece of silicon, aircraft-sized if necessary, where the storage units are not transistors but physical cells. Nothing about that is alive, so nothing about it can drift. It costs a building and a power supply, and it will hold a century if the building holds.
The honest comparison is not against a hard drive. It is against the question a century actually asks: not whether the bits survive, but whether anyone in a hundred years can read them. A genome survives perfectly and is meaningless without the sequencer, the encoding scheme and the chemist.
So the durability is easy and the legibility is the problem. Any archive measured in centuries is a bet that the reader will exist.