Your Bitcoin wallet can generate a new address in seconds. It cannot change the fact that 0.73184627 BTC is a very specific number.
Move something close to that amount elsewhere, at a revealing time, through a recognizable transaction pattern, and the number can become a clue. Change the address again. Add another transaction. An observer may still have enough information to connect the dots.
That is one of the uncomfortable realities of Bitcoin privacy: an address can be disposable while an amount remains memorable.
For years, privacy developers have worked to weaken the relationships outsiders infer from transactions. Babilonia v2 takes that effort in an unusual direction. It explores whether real, privately negotiated bets can change the amounts participants receive enough to make familiar matching techniques less reliable.
The blockchain still records the correct numbers. The bet changes the economic event behind them.
On September 5, 2026, Adam Gibson announced version two of the Babilonia paper and a browser demonstration called Thimbly, running on Bitcoin’s Signet test network. The public discussion began in July and includes the author’s discovery of a revealing weakness in the original design. That weakness prompted the redesign. ** Gibson’s development thread** documents the sequence.
The story deserves attention because it asks a question that reaches beyond gambling:
How much can surveillance actually learn from a perfectly accurate ledger when the economic relationships underneath it become harder to predict?
That question matters to anyone who wants to use Bitcoin without publishing an easily reconstructed financial biography. It also comes with a difficult trade-off. The uncertainty Babilonia creates involves actual value changing hands. Someone bears the risk.
The interesting possibility is stronger privacy. The difficult part is deciding what that privacy costs, whom it protects against, and whether it survives contact with an adversary.
Washington can repay every dollar and still return less of your future. Gold’s greatest threat is exposing the difference.
Imagine withdrawing bitcoin, moving it through several transactions, and eventually making a payment. Each step produces fresh addresses. From the wallet screen, everything looks separate. From the blockchain, the transactions still sit inside a public graph.
A transaction spends earlier outputs and creates new ones. Those outputs have explicit values. Anyone inspecting the chain can follow the spending relationships, see when transactions confirm, and study the amounts involved. Bitcoin.org’s privacy guidance also stresses that information which is difficult to interpret today may become easier to connect later. ** Bitcoin.org’s privacy guidance** describes this permanent exposure.
Identity enters through additional information. A merchant knows who placed an order. An exchange knows which customer requested a withdrawal. Someone publishes a donation address. A database leaks. An investigator combines those observations with the transaction graph.
Amounts can help connect otherwise uncertain relationships. Suppose an observer is looking for the destination of roughly three quarters of a bitcoin. A nearby transaction carrying nearly that value may attract attention, particularly when its timing and surrounding activity fit the hypothesis.
That is an inference. Similar amounts occur for unrelated reasons, and good analysis must account for that. But a clue does not have to be conclusive to be useful. It only has to reduce the number of plausible explanations.
Fees complicate the comparison without necessarily destroying it. If the expected difference is small, the observer can search within a narrow range. Splitting funds changes the shape of the problem, but combinations of outputs may still reveal a meaningful total.
This is why privacy cannot be reduced to a single instruction about addresses. Address hygiene addresses one source of exposure. The amounts, timing, transaction structure, and information held by counterparties remain separate questions.
Babilonia concentrates on one of those questions:
What happens when the amount that emerges from an interaction genuinely differs from the amount an observer would normally expect?
There is a gap between what Bitcoin verifies and what an observer wants to know.
A full node checks whether a transaction satisfies the consensus rules. Among other things, it checks that the inputs are available to spend, that authorization is valid, and that the transaction does not create unauthorized value. It does not need to decide whether the people involved are buying coffee, moving savings, settling a debt, or closing a private agreement.
Those economic descriptions belong to another layer of interpretation.
Even the assumption that every input in a transaction belongs to the same person can fail. PayJoin provides a concrete example: a receiver can contribute inputs to a payment transaction alongside the sender. The transaction remains valid, while the common-input ownership heuristic becomes unreliable in that case. The PayJoin specification explicitly identifies this and other heuristics as targets. BIP 78 explains the design.
The implication extends beyond that particular technique. A public transaction can support several possible economic explanations, even when every byte of the transaction is known.
Picture two friends settling several obligations at once. One owes money for dinner. The other owes part of a holiday booking. They agree on a single final transfer. A third party can see the settlement amount without learning the full set of obligations that produced it.
Bitcoin does not require an invoice for every economic reason behind a payment. That leaves room for privacy through how people organize their transactions and agreements.
Babilonia explores a particularly unusual agreement: a probabilistic transfer whose result is enforceable using Bitcoin. The participants privately establish the rules, then settle the outcome. An outsider sees the resulting transactions but may lack the information needed to reconstruct the underlying allocation of value.
A correct record of settlement is not automatically a complete record of the relationship being settled.
That distinction is where the experiment begins.
The word “bet” can make Babilonia sound like another casino application. That reaction is understandable. It also overlooks the reason a privacy researcher might care about betting in the first place.
