Field note · 2 February 2026 · 8 minutes
The custody ladder
Where a coin sits is a decision about failure, not about taste. Each rung should have a job you can say without using the word "easy."
Custody arguments online tend to collapse into loyalty. One camp thinks leaving coins on an exchange is naïve. Another thinks self-custody is a way to lose a fortune to a piece of paper in a drawer. Both camps are describing a real failure, and then pretending it is the only one. A ladder is how this desk refuses the pretend choice. You name the failures you are willing to carry at each step, and you stop asking a single rung to be wise about all of them.
Rung one: the working exchange
The first rung is a venue you use because you might need to act in days, not because it feels like a bank. It is the wrong home for the whole sleeve. Withdrawals can pause. Accounts can be frozen while a firm checks its own paperwork. A balance on a screen can be a claim on the firm rather than coins you control. The position-band note caps this rung, in its illustration, at 40 percent of the sleeve. The percentage is invented. The category is not: money that has to move soon should be small, and it should be assumed to be stuck at the worst moment.
We do not name a preferred venue. A name would go stale, and a recommendation is not our work. What belongs in your notebook is the reason that venue is allowed to hold anything at all, the ceiling, and the date you will check whether withdrawals still work. "They have an app" is not a reason.
Rung two: a second door
A second venue is not diversification of ideas. It is a second door. If the first firm spends a fortnight in maintenance, or worse, you still have a way to reduce risk or raise sterling without begging a support desk. The second door should be set up before you need it: identity checks finished, a tiny test deposit and a tiny test withdrawal already done, records kept. Doing the test during a panic is how people approve a phishing screen.
Two doors do not make either firm safe. They make a single firm's bad week less than a total halt. That is a modest goal, and modest goals are the ones frameworks can keep.
Rung three: self-custody for the long sleeve
The long sleeve, if you have decided one exists, is the money you have agreed not to touch because of a headline. Putting that money where a firm can halt it contradicts the agreement. Self-custody is the rung that matches the job, and it brings a sharper failure: if you lose the secret and the backups, nobody has a duty to restore it. There is no ombudsman for a forgotten drawer.
This note will not show you a seed-phrase layout, a photograph of a recovery card, or a sequence of buttons. Those details help thieves as readily as they help readers, and they age badly as devices change. What we will say is structural.
- The backup lives apart from the device, in a form that survives fire and a casual visitor, and it is not stored in a cloud photo album.
- A second person may know that an envelope exists and who to call. They do not know the secret unless you have made a deliberate, written choice that they should, and you have accepted that choice as a risk.
- The device you use for the long sleeve is not the device you use to try new applications on a bored evening.
- You can explain, without notes, what would happen if you were in hospital for a month. If you cannot, the rung is unfinished.
The drill
Once a year, on a date written at the front of the notebook, restore a tiny amount to a fresh setup and send it back. The amount should be small enough that a mistake is a lesson in sterling you can name without going quiet. The drill is not a performance. It is how you discover that a backup you have never tested is a rumour.
If the drill fails, you do not "solve it later." You stop adding to that rung until the restore works. Adding to a custody path you cannot exit is how a framework becomes a trap with good typography.
What we leave out on purpose
No brand ranking. No affiliate link. No instructions that would help someone empty a wallet they do not own. If a letter arrives asking us to publish those missing steps "for safety," the answer is no. The structural page is the safety we are willing to put in public. The rest belongs in the manual of the tool you chose, read slowly, offline if you can, and never from a link in an unexpected message.
Custody does not make a bad thesis good. It only decides which kind of loss you are underwriting. Read it next to the risk discipline page. And remember the standing line of this site: we are not your custodian, your broker, or your adviser.