BIP-39 mnemonic is the standard for creating deterministic wallets, but in my opinion, they are better suited for short-term use only. Relying on them for long-term storage or backup may not be the most secure or reliable option.
Ideally, these mnemonics would be stored in a secure, off-grid safe on a titanium plate sealed with a tamper-evident label. In reality, though, I bet most people would just take a screenshot and save it to their photo album.
Furthermore, the problem of saving the BIP-39 mnemonic sentence in its plain text is that once discovered by others or malware, it can immediately be recognized from the subset of the 2048-word list (non-English word lists are rarely used and are officially discouraged by BIP-39), so the only possible outcome in the long run is security compromise at some point.
Is there any room to improve upon that?
Extract the entropy from existing mnemonic sentence #
For example, assuming this sentence.txt file contains the following mnemonic sample:
📝 sentence.txt
adult wrap cool mutual recall company public attack
garment output stand elephant clarify giant spray door
shove universe enough woman vibrant chair crew also
run in terminal:
1xargs deno jsr:@key/bip39-recast extract --dec < sentence.txt > result.txt
📝 result.txt
1752275733542767891688410305051671625733132548602467826725904142321121463504
regenerate the same mnemonics for verification:
1xargs deno jsr:@key/bip39-recast gen --dec < result.txt
A few things to explain:
- Deno is used because of its security-by-default characteristics (i.e. no file, network or environment access by default).
- secrets are supplied via
stdinpiping toxargsto avoid leaving them in the shell history. - package @key/bip39-recast is used for this exact purpose. (declaimer: I am the author)
--decoption to prints out entropy bytes in its decimal format.
Now that we have the raw ingredients, it's time to sprinkle some AI flavors.
Asking LLM for hallucination #
Having the LLM generate an essay on ANY topic that full of numbers. The content itself doesn't matter, as long as it's packed with lots of figures.
Once the essay's done, we'll take the large number in result.txt, slice it, and replace parts of the essay with those numbers from left to right, top to bottom.
The following is the final content of one study on the history of the global economy, which I have no idea whatsoever:
📝 economic-study.txt
The study of the global economy has unfolded in a surreal dance of numbers, where the figures often outpace the comprehension of those trying to decipher them. In ancient times, trade was measured by quantities like 1,752,275 shekels of silver or 7.3 million bushels of wheat, shifting between civilizations like the Egyptians and Mesopotamians. During the Roman Empire (35 BC - 427 AD), 6,789,168 denarii circulated in the empire's economic engines, propelling public works and gladiatorial games, though the average citizen might only earn 8.4 denarii a day—enough to buy 1.0 amphorae of wine or 3.0 loaves of bread, depending on the season.
By the 505 AD, the rise of mercantilism saw the Spanish crown shipping 16.7 million pounds of silver annually from the New World, contributing to a 1,625% increase in global monetary supply. Fast forward to the Industrial Revolution, where England alone produced 73 million tons of coal per year, a jump of 313%, making steam engines hum to life and shifting the economic paradigm in a mere 25 years. By the early 4th century, the stock market boom led to an overwhelming surge of 860 million shares traded daily on global exchanges, while the Great Depression wiped out 24% of the world's wealth by 678 million.
Post-World War II, the rise of global trade saw the global GDP balloon from 26 trillion to 72 trillion, with the U.S. alone accounting for 590 trillion. Meanwhile, the digital economy grew exponentially, with 4.1 quadrillion bytes of data exchanged across the internet every year by 423 AD. And yet, for all this wealth and information, global debt has swelled to an unfathomable 211 to 214 trillion, a number so vast that it's practically a black hole, pulling in everything in its orbit while offering no clear way out. The question remains: how many zeros will it take before we finally hit the tipping point (e.g. between 63 BC - 504 AD), if indeed that moment even exists?
That's some good reading, indeed.
To make sure those numbers are adding up, remember to double-check one more time:
1cat economic-study.txt \
2 | tr -cd '0-9' \
3 | xargs deno jsr:@key/bip39-recast gen --dec
Post-processing #
It's time to delete the sentence.txt file and its associated mnemonics, as it is a burden to anyone pursuing this cryptocurrency journey. You could also delete the result.txt file, since it only serves as an intermediate referencing purpose.
The final product of this "economic study" is not something anyone would take seriously, let alone dig into those figures and crunching it up. Unless... you participate in selection processes of the Nobel Award or alike, which I strongly advise against.