How good a record is good enough?

No record in English (and I'm using English as a shorthand for any human language) can ever be good enough. English is not a technology for transmitting information.

English is a compression format, and a very lossy and somewhat inaccurate compression format at that. But it has a stupendously high compression rate and compression algorithms with reasonable running speeds on specially adapted hardware (i.e. brains), so for the particular purposes of human communication English is a pretty decent option.

I own a t-shirt with this graphic printed on it. If you possess a mostly correct compression algorithm (that is, you speak modern English), the ~5kb of data on that shirt contains sufficient information to reproduce ~30 major scientific or technological discoveries. I don't know exactly how much space you could fit that information into if you encoded it in a way that wasn't very heavily optimized for very specific types of human brains, but I suspect it's many orders of magnitude greater than 5kb.

On the surface this seems like it could be an argument for reproducing a specific human from their preserved written material, what with the amazing information density of English. But using a standard English decompression algorithm to analyze what you've written is worthless, because we're not trying to recreate the meaning of what you've written. We're trying to recreate the compression algorithm used to create your writings, which would be approximately isomorphic with your brain. But because the English data format is lossy and imprecise, reconstructing that algorithm from only its output is impossible.

If you could preserve a copy of the decompressed version of what you were trying to write along with your writings, that might be enough to reverse-engineer your brain('s compression algorithm). But I don't think that's possible for any human, much less most of them.