Two pre-Bitcoin proposals left one problem unsolved

Inherited archive entry · retained from the earlier edition. Its factual, legal and policy claims have not all been reverified for 19 September 2026. Research scope.

Two designs, written years apart, that describe most of the parts

Wei Dai's b-money proposal, posted to the cypherpunks mailing list in November 1998, sketches a system where every participant maintains a separate database of how much money belongs to each pseudonym, updated by broadcasting signed messages to everyone. Nick Szabo's 'bit gold' post describes a different route to a similar goal: a chain of proof-of-work puzzles, each timestamped and linked to the one before, with ownership recorded in what Szabo calls a distributed property title registry, so that no single server is substantially relied on. Both texts assume computational cost can substitute for a mint's authority to create scarcity.

The step both proposals leave open

Szabo is candid that the security of his scheme depends on how well trust can be distributed across the timestamping and title-registry steps, without specifying how a registry made of many servers reaches one agreed answer when they disagree. Dai's first b-money variant needs every participant to keep a synchronised copy of every account, which does not scale past a small, cooperative group; his second variant delegates bookkeeping to a subset of bonded servers, which reintroduces a smaller trusted set rather than removing one. Neither document specifies a rule by which mutually distrusting strangers, most of whom have never met, converge on a single ordering of events without appointing someone to referee it.

What Bitcoin's whitepaper added, by its own account

Satoshi Nakamoto's paper lists Dai's b-money as its first reference and proposes the missing rule directly: nodes accept the longest proof-of-work chain as proof of what happened while they were gone, so agreement is produced by costly computation rather than by appointing a registry operator. That single mechanism, a shared chain extended by whoever expends the most verifiable work, is the specific gap neither the 1998 nor the later proposal closes.

A note on the evidence itself

One complication is worth flagging plainly: the bit gold post's own web address dates it to December 2005, the month widely cited in later histories, yet the page currently displays a 2008 date banner above the title. This is likely an artefact of the blogging platform rather than a genuine publication date, and it means the essay's exact date cannot be stated with confidence from the document alone; no specific day for it is asserted here.

• When two designs agree on every mechanism but one, is the missing mechanism the real invention?
• Does distributing trust across many servers remove a trusted party, or just multiply it?
• What happens to a system's history when its own supporting documents cannot be dated reliably?

Reading b-money and bit gold beside the whitepaper narrows the claim of originality to a specific, checkable place: not the idea of digital scarcity, but the rule for agreeing whose copy of history is correct.