Writing · October 2026
The dice were loaded.
A seed phrase is supposed to come from a roll of dice nobody can see. This week, for a lot of wallets, the dice were loaded, and someone noticed.
Qubic halted its network on Sunday, October 18, after an attacker drained wallets whose seeds were generated predictably. Crypto Briefing reports 1,404 wallets emptied between 23:42 UTC on October 17 and 01:26 UTC on October 18, about an hour and forty-four minutes. Validators stopped the chain at tick 85,371,949. I'm not writing about the project. I'm writing about the failure, because it's the same one as the Coldcard exploit this summer, in a different place.
What the reporting says
Every number below traces back to Qubic's own statements. Cointelegraph says it did not hear back from Qubic and did not independently verify them. Treat them as the project's account, not a settled record.
Qubic says about 2.94 trillion QU was taken. It says the cause was weak seed generation, not a flaw in the network's core code. The affected seeds, per Cointelegraph, were those created by any browser on an iPhone or iPad, by Safari on a Mac until October 18, or by the Qubic iOS app before July 2025. Cointelegraph notes it is unclear whether that Mac date marks the end of the weakness or the date it was fixed. That ambiguity matters, and I'd want it answered before trusting anything.
On the fix, Cointelegraph reports a GitHub commit for version 4.26.4 that replaces Math.random() with crypto.getRandomValues() in seed generation. The first is a general-purpose function and not designed for secrets. The second is built for them. That detail comes from the commit as the outlet describes it, not from a full Qubic post-mortem, which the project says will come after a recovery plan is settled.
On the money, Qubic says 87.1% of the drained funds never left its network and are frozen in place. The other 12.9% went to the exchange MEXC, which Qubic says froze the receiving account within about an hour. Crypto Briefing adds that Bitget suspended deposits and withdrawals for the token. Network validators, called Computors, will vote on what happens next.
Where outlets disagree: the dollar value. Crypto Briefing and The Defiant's headline say about $1.3 million. Cointelegraph says about $1.4 million. No timestamp is given for either, so I'd call it roughly $1.3 to $1.4 million and leave it there.
A patch can't repair a seed that already exists
This is the part I'd hold onto. A seed is generated once. After that, the generator doesn't matter, because the seed is just a fixed string, and if the process that produced it was guessable, so is the string. Updating the software changes what the next wallet gets. It does nothing for the ones already made.
I wrote about the same shape in August, when a Coldcard firmware bug reduced effective key strength for some seeds. Cointelegraph draws the same line, pointing to that case and to the 2023 Trust Wallet extension and Libbitcoin Explorer weaknesses. Different vendors, different years, same root: randomness that wasn't.
Think of a square. If the square is off by a hair, every cut you made with it is off by a hair. Buying a corrected square doesn't fix the frame you already built.
What I'd check on my own wallets
- Know where each seed was born. Hardware device, browser page, desktop app, phone app? Write it down. If you can't say, you have a gap to close.
- Treat seeds made in a browser or an old mobile app as the ones to review first. Not because every one is weak. Because the place a seed is made is the place an attacker looks.
- If your seed's origin is in doubt, make a new one on a device built for it and move funds across. Send a small test amount first. Don't reuse the old phrase for anything after.
- Read the vendor's advisory from a page you bookmarked on a calm day. Not from a post or a message. Incidents like this are exactly when fake "migration tools" show up, which is the subject of my last piece, The Dice Were Loaded.
- Never type a recovery phrase into a website to check it or migrate it. Nothing legitimate needs that.
The boring conclusion
The dice either are random or they aren't, and you can't see the roll. All you can do is control where it happens and how soon you find out it was bad. Ask the question about your own seeds while nothing is wrong.
Verify everything, including how your keys were made.
— Jon Tetreault, October 2026. The habits, tools, and checks behind this piece are what I teach in Crypto Security Mastery.