Why Seed Phrases Were a Mistake

Ask someone why they don’t hold their own crypto and you’ll rarely hear “I don’t understand the economics.” You’ll hear some version of: I’d lose it. They’re right to worry. We built an entire financial paradigm on top of the assumption that ordinary people can safeguard twelve random words forever, and then acted surprised when they couldn’t.

Where the twelve words came from

The seed phrase is not ancient. It was formalised in 2013 as BIP-39, a Bitcoin Improvement Proposal, and its stated purpose was straightforward: a private key is a 256-bit number, and 256-bit numbers are miserable for humans to write down, read aloud or type without error. Encoding that number as words drawn from a fixed 2,048-word list made it transcribable. Checksum bits meant a mistyped word could be detected rather than silently producing a different, empty wallet.

As an encoding scheme, it’s elegant. That is genuinely what it was designed to be - an encoding scheme, replacing something worse.

What it was never designed to be is the primary consumer-facing security model for a global financial system used by hundreds of millions of non-technical people. It became that by default, because it was there, and because for a long time everyone using crypto was technical enough to cope with it.

The failure mode is the problem, not the difficulty

Plenty of security systems are hard to use. What makes seed phrases unusual is the shape of the failure.

Consider what happens when you forget your online banking password. You click a link. You receive an email. You are inconvenienced for ninety seconds. The system was designed on the assumption that humans forget things, because humans forget things.

Now consider what happens when you lose your seed phrase. There is no link. There is no email. There is no institution with a record. The money is not frozen, it is still visibly there, on-chain, addressable, and utterly unreachable. It stays that way forever.

And the phrase has to survive an implausibly long time. Not until next Tuesday, indefinitely. Through house moves, floods, fires, divorces, deaths, and the ordinary entropy of paper in drawers. Meanwhile every instinct people have developed for keeping information safe photograph it, email it to yourself, save it in your password manager, tell your partner, which converts the phrase into something a thief can reach.

So the advice becomes a list of prohibitions. Don’t photograph it. Don’t type it. Don’t store it digitally. Don’t keep it near the device. Don’t tell anyone. Also, don’t lose it, ever, or your money is gone.

That isn’t a security model. It’s a set of instructions for not being a normal person.

“Users just need to be educated” is not a design

For roughly a decade, the industry’s answer to this was better onboarding. Clearer warnings. Forced confirmation screens where you re-enter words four, seven and eleven. Metal plates stamped with your phrase and buried in the garden.

Some of this helped. None of it addressed the actual problem, which is that the system’s correctness depends on flawless human behaviour sustained over decades, with catastrophic and irreversible consequences for a single lapse.

When a system fails constantly in the same way across millions of independent users, the users are not the variable. Aviation worked this out in the 1970s: if trained professionals keep making the same mistake, you redesign the cockpit rather than writing another memo. Software worked it out too which is why nobody today ships a product where forgetting your password destroys your account.

Crypto kept writing memos.

The custodial answer, and why it isn’t one

There has always been an obvious escape: let someone else hold the keys. Exchanges and custodial wallets do exactly this, and for many people they work fine, right up until they don’t.

The trade you’re making is specific and worth understanding properly. You swap “I might lose this” for “someone else might lose this, freeze it, be hacked, be compelled by a court, or go bankrupt.” The last decade supplied ample evidence for every item on that list. Custody solves the memory problem by reintroducing precisely the counterparty risk that self-custody existed to eliminate.

So the real question was never “seed phrase or custodian.” It was whether you can get key ownership without a single fragile secret.

What replaces the seed phrase

You can. The insight is that a key doesn’t have to exist in one place to work.

Social recovery distributes the ability to restore access across parties you choose, trusted contacts, your own second device, an identity provider you already use daily. No single one of them can move your funds. A quorum of them can help you regain access. The concept was popularised for smart-contract wallets by Vitalik Buterin in 2021, and the design goal is stated plainly: eliminate the single point of failure without creating a custodian.

Multi-party computation goes further. The full private key is never assembled anywhere, not on a server, not on your phone, not even momentarily. Independent shares cooperate to produce a signature. There is no complete key sitting somewhere to be stolen, and no phrase to lose, because the thing a phrase would encode never exists as a single object.

Both approaches share one property that matters more than their cryptographic differences: losing one thing is survivable. Lose your phone, keep your account. That is how every other system you depend on already works, and it is not a compromise on self-custody in both models, the provider still cannot unilaterally move your funds.

Seed phrase vs social recovery

A closer look at how the two models differ, what each one actually protects against, and where social recovery has real trade-offs of its own.

→

The trade-offs are real

It would be dishonest to present this as free.

Social recovery means choosing guardians, and that choice has social texture people move, fall out, die, or can be manipulated into approving a recovery they shouldn’t. Recovery flows are a genuine attack surface, and a badly designed one is worse than a seed phrase. MPC implementations vary enormously in quality, and “MPC” on a marketing page tells you nothing about whether the scheme was audited or whether one share-holder is quietly load-bearing.

The honest claim is narrower than “this is strictly safer,” and it’s this: seed phrases optimise against the wrong threat. They’re excellent at resisting a remote attacker and catastrophic against the thing that actually separates people from their money, which is ordinary human forgetting. A model that survives forgetting, and that you can reason about with the instincts you already have, protects far more people in practice even if a determined attacker has more surfaces to probe.

Why this is the whole ballgame

Every projection about crypto reaching mainstream use quietly assumes people will hold their own assets. Self-custody is the entire point, it’s what distinguishes this from a slower, stranger banking app.

But you cannot ask a billion people to accept a system where one lapse of memory is financially fatal. They will decline, correctly. They will use a custodian, or they’ll use nothing, and either way the thing the technology was for doesn’t happen.

Seed phrases were a reasonable engineering decision in 2013 that hardened into an assumption nobody revisited. They weren’t a mistake because the cryptography was wrong it wasn’t. They were a mistake because a transcription format was promoted into a consumer security model, and then defended for a decade on the grounds that users should try harder.

They shouldn’t have to. That’s what we’re building at DYNK: a wallet you get back by signing in again with an account you already control, not by finding a piece of paper, and no twelve words standing between you and everything you own.