I write bitcoinkeys.guide. For months it led with choose the simplest setup that covers you, and I believed it — complexity you don't fully control has buried more coins than thieves have ever taken. Nobody in this room needs convincing of that.

Then 30 July happened. Someone emptied hundreds of wallets in under an hour, because a build-configuration error from March 2021 had quietly made one hardware wallet's seed generation fall back to a weak RNG — roughly 40 bits of effective entropy instead of 128, on the worst-affected models. It sat in public, open-source firmware for five years, in the single most security-critical function a signing device performs. The audits didn't catch it. Reproducible builds didn't catch it. I didn't either: I rated that device in my cold-storage tier and recommended it in my own setup finder, which is why I'm not going to be quiet about it. To be fair to the maker, a unit bought today on fixed firmware generates a proper seed — the defect is in seeds already created, not in every device forever.

Every wallet drained that day was single-sig with no passphrase. The owners who came through weren't running better hardware. They'd combined something — a passphrase the device never saw, dice they rolled themselves, a key from a different maker. Any one of those and the flaw couldn't reach them on its own. Nobody predicted that specific bug, and that's exactly the argument: you can't defend against a threat you haven't imagined, but you can refuse to own a component whose failure is total.

So I demoted simplicity. It's still the target — it just stops winning the moment it would leave one thing holding everything. My reasoning is that most of the guidance this field rests on was written when attacks were expensive, hand-made and rare, and they're now cheap, automated, and increasingly assembled by machines that read code faster and more patiently than any auditor. And to head off the obvious objection: this is not a multisig pitch. One seed plus a passphrase that device has never seen is two independent things and clears the bar. Three keys from one manufacturer in one batch don't — that's one thing bought three times. The whole position, with what would change my mind: https://bitcoinkeys.guide/about/

Here's where I actually want the argument. The strongest case against me is that I'm trading a rare, exotic failure for a common, boring one — passphrases get forgotten, second factors get lost, and every extra component is one more thing to keep right for years on your worst day. That's a real cost and I might be underweighting it.

So: is simplicity-first still the right default advice for a normal holder, and am I overreacting to one bad firmware year?