Test Your Backup Before It Matters

If you haven't tested your Bitcoin backup, you don't have one. You have a piece of metal with words on it that you hope will restore your funds when everything depends on it. Hope is not a recovery strategy.

A firmware update cannot retroactively improve the randomness used to create an existing seed. A recovery drill checks whether a backup restores the expected wallet; it does not prove that the seed was generated securely. These are separate questions.

The recovery drill

This is a reproducible procedure for verifying that your Bitcoin backup actually works. You execute it yourself, on a scratch wallet with a trivial amount. An observer witnesses and records the result. The observer never sees or touches your keys.

Step 1: Create and fund a scratch wallet

Create a new wallet on your hardware device. Send a small amount to it, enough to test with. Five dollars is sufficient. This is not your real wallet. It exists only to be destroyed.

Step 2: Wipe it

Delete the wallet from the device. Clear it completely. The funds are now accessible only through the backup you recorded when you created it.

Step 3: Restore from your real backup

Use your actual backup materials, the seed phrase, the metal plate, whatever you stored when you first set up the wallet. Import it into a fresh wallet on the device or in a software wallet. Type it in exactly as you wrote it.

Step 4: Prove the funds are spendable

Send the test funds to a different address. If the transaction broadcasts, your backup works. If it fails, you have found the problem now, when the stakes are five dollars, not when the stakes are everything.

Step 5: Wipe again

Clean the scratch wallet completely. You have proven the recovery path works. The test is complete.

Step 6: Record the result

Record whether the drill passed or failed, which steps worked, which failed, and what was corrected. This record is the evidence that your backup has been verified. It does not contain any secret material.

What the drill catches

Three failure modes are the most common:

• Transcription errors. A seed phrase written down months ago with a wrong word, a missing word, or words in the wrong order. The checksum rejects it, or it restores to an empty wallet.
• Wrong derivation path. The seed imports correctly, but the recovery wallet defaults to a different address format. The restored wallet shows zero balance. During an emergency, you might assume your funds are lost.
• Passphrase confusion. You forgot you set a passphrase, or you misremember the exact string. Recovery without it produces an empty wallet. Recovery with a slightly wrong passphrase also produces an empty wallet.

Non-custodial by construction

The observer's role is to witness and document. The observer never requests, receives, generates, or stores a seed phrase, passphrase, or private key. The observer never sees your real wallet balances. The drill runs on a scratch wallet with trivial funds. This is not a policy. It is a structural impossibility built into the procedure.

When to run it

Run the drill when you first set up a wallet. Run it again when you add significant funds. Run it when you change hardware, when firmware updates ship, when you change heirs, or when enough time has passed that you are no longer sure it would work. Once a year is not excessive.

If the conditions under which a seed was generated are uncertain, a successful restore does not establish that the seed is secure. Treat seed-generation quality and backup recoverability as separate checks. Never assume that updating firmware or passing a recovery drill resolves both.

I would rather learn my backup is broken on a Tuesday with five dollars at stake than on the worst day of my life with everything at stake.

Run the drill with an observer

You can run every step above on your own, and you should. If you want a second set of eyes, ask Loop XXI about the scope and availability of educational support. Any agreed observation must exclude recovery words, passphrases, private keys and wallet PINs from screen share.

The observer never requests, receives, generates, or stores a seed phrase, passphrase, or private key, and never takes custody of funds. Inquiry-only; scope and rate on request.

Request a recovery drill →

Pay for AI with the same sats

You already hold bitcoin. You can also spend it on AI inference without an account, without a subscription, and without handing over your email. Loop Gateway, built and operated by Loop XXI, meters API usage against prepaid credit at upstream token cost plus a 15% markup. Credit can be purchased with Lightning; each API call consumes that balance. The same sats that protect your backup drill can pay for a research query, a code review, or a summarization task.

Buy inference credits with bitcoin →