Two Attempts Away from Digital Silence

In 2011, Thomas created a short animation called What is Bitcoin? He was paid in bitcoins and after a time retained 7,002 BTC. It was not a secret wager on billions; their value then was small compared with today’s calculations, and the young network was looking for people who could explain it clearly in the first place.

He had the keys in three backups. According to his later account, he accidentally deleted two. The third was stored on an IronKey S200, an encrypted USB drive with a limited number of password attempts. He wrote the password on paper. The paper disappeared.

Three copies sound like sensible redundancy, but all depended on the same human layer. Two could be deleted and the third opened only with information kept on a single slip. The number of backups alone does not tell us how many genuinely independent recovery paths exist. Three technical media can share one point of failure.

The video What is Bitcoin? lasted barely a minute and a half. Colored figures sent digital money directly to one another while a simple voice explained properties that now fill entire books. Thomas was not being paid for a financial prediction; he was paid to translate a technical idea into human language. The coins were a fee for work performed when one bitcoin cost less than a dollar, not a ready-made fortune that a reasonable person would automatically store as a family treasure.

The later mistakes become unbearable precisely because the scale changed. When Thomas told WIRED in 2011 that he had spent a week trying to recover access, the balance was worth about 140,000 dollars — painful, but still conceivable. In later years its value repeatedly rose into the hundreds of millions. The same piece of hardware, the same forgotten password and the same 7,002 BTC became an entirely different psychological problem without a single bit on the device changing.

Thomas said in interviews that he had entered eight wrong passwords out of ten. Two attempts therefore remained before the device was supposed to make the data inaccessible. “Two left” is a state described publicly, not live telemetry from the device; without confirmation from Thomas, no one can know whether it later changed.

The balance’s value kept rising, and with it the psychological cost of pressing Enter. The ordinary advice to “try what you use elsewhere” becomes a gamble when a mistake consumes one of a finite number of lives. Thomas described periods when he would wake with a new idea, try it and fall back into helplessness.

The limit also changes how a person thinks. Without one, a list of candidates can be tried systematically; with two attempts, he must first reconstruct his own former mind. Did he use a phrase, a random string, an unusual spelling? Every new memory may be either the breakthrough or false certainty. The device does more than protect data from an attacker — it forces the owner to prove his identity without forgiving ordinary human imprecision.

Thomas described a cycle familiar to anyone who has searched for a lost password, only with unimaginable stakes. At night a new combination would occur to him; he would get up, walk to the computer and briefly believe that this time it was right. Failure brought not merely a technical error but despair and a return to bed. After eight attempts, another inspiration was no longer a gift; it was a question of whether it deserved to consume half the remaining hope.

A number quoted by an exchange encourages the image of a pile of dollars waiting behind the screen. The IronKey held neither dollars nor bitcoins themselves. It held an encrypted file that could enable a signature. Until Thomas opens it, he cannot sell the coins, pledge them or move them into a safer arrangement. The wealth can be priced publicly and is simultaneously unusable — an arm’s length away, yet one correct memory out of reach.

In 2023, the Unciphered team demonstrated an attack on the same IronKey model to WIRED. On a test device, it bypassed the ten-attempt limit and found the correct three-word password after an enormous number of guesses. The company believed that its method could unlock Thomas’s drive as well.

Thomas did not give them the device. He said he already had agreements with other teams and that negotiations over such a high value take a long time. The story therefore has no confirmed happy ending. There is a demonstrated technique on a comparable device, not public proof that Thomas’s keys were recovered.

Unciphered did not describe a magical universal key. Its work involved a laboratory attack on a particular hardware generation, extracting internal data and bypassing the mechanism that counted failed attempts. Such an attack can require time, equipment and trust in a specialist holding a device of enormous value. Technical possibility is therefore not yet a finished service or a safe deal.

The team named the project Everest. Chief operating officer Nick Fedoroff, chief executive Eric Michaud and a hacker using the name Tom Smith collected hundreds of old IronKeys, X-rayed chips, etched away protective layers with acid and reconstructed the device’s internal logic. When Smith found WIRED reporter Andy Greenberg’s three-word password on a test unit after roughly 200 trillion attempts, he proved something very specific: under laboratory conditions, the counter in this generation could be bypassed.

The hardest barrier was no longer silicon but a contract. Thomas already had verbal agreements with other experts, including the company Naxo and researcher Chris Tarnovsky, according to WIRED, and declined to begin another negotiation. With hundreds of millions at stake, an owner must resolve the fee, liability, copies of the data and the possibility that the rescuer also gains access to the key. Fedoroff summed it up without technical vanity: code and circuitry behave predictably; people do not.

The story is often told as the failure of a hardware wallet. An IronKey, however, is general-purpose encrypted storage. It did not securely sign Bitcoin transactions like a modern dedicated wallet; it protected a file containing keys with a password and an erase policy. The distinction is more than wording because it leads to a different lesson.

Security is not the strongest possible lock. It is a system in which a person can recover the secret, heirs can understand it and one mistake cannot destroy every route. Thomas’s IronKey resisted strangers extremely well. The tragedy arose because it resisted its rightful owner with the same severity.

Modern wallets address part of the problem with a standardized seed phrase and the ability to recover keys in compatible software. That alone is still not enough: badly stored words can be stolen, destroyed or inherited without instructions. The point is not to remove responsibility, but to distribute it so that the loss of one object or one memory is not a final sentence.

Thomas’s case also exposes the weakness of simply ranking security from weak to strong. IronKey had certified resistance to tampering, and its development was partly funded by the U.S. Department of Homeland Security. The device’s resilience was not itself a flaw. The problem arose when the threat model changed: from the device’s perspective, the authorized owner who did not know the password became an attacker. The individual device was well protected, but the overall arrangement for retaining access proved catastrophically fragile.

The human point is not that Stefan Thomas should have remembered better. Everyone’s memory fails, and paper can disappear. The point is that a custody design must account for that ordinary fact before value or illness turns recovery into a crisis. Two remaining chances on an IronKey are more than a countdown to a technical lockout. They are a reminder that sound ownership includes a road back to one’s own key.