A Bitcoin developer has demonstrated a way to check the network’s entire spendable supply in roughly the time it takes to copy a large file, an experiment that hints at how much lighter a full node could become.
The work, published to the Delving Bitcoin forum on Nov. 13, used a snapshot of about 164.5 million unspent transaction outputs — the records of bitcoin received but not yet spent — captured at block 972,500. The file weighed 9.39 GB. On an arm64 machine with roughly 0.5 GB of memory, checking it against a cryptographic proof took about 72 seconds, nearly all of it spent reading from disk.
Separately, the developer showed that a compact 1.6 MB proof could be verified in a web browser in about half a second, raising the prospect of a client that checks Bitcoin’s ledger state without holding it.
The problem being attacked is setup time. A fully validating node rebuilds the set of spendable coins by processing every historical transaction, which is why running one has drifted toward dedicated hardware and patience. Bitcoin Core already offers a shortcut called assumeUTXO, which lets a node start from a snapshot and verify history in the background. The experiment aims at the verification itself: swapping a re-run of years of transactions for a STARK proof that the snapshot matches the chain’s commitments, since these proofs let a machine confirm a computation without repeating every step.
The boundaries are narrow, and the author states them plainly. The proof covers block headers from Bitcoin’s genesis and checks changes to unspent coins between blocks 935,001 and 972,500. Everything before that range is not replayed; the node trusts a starting snapshot and verifies a bounded stretch forward from it. The two timing figures also measure different things — one is a check of the 9.39 GB file against the proof’s commitments, the other a browser check of the 1.6 MB proof itself.
What the check establishes is that the coin set matches those commitments: base block, count, amount and hash. What it omits is the part of validation that costs the most in practice. Scripts and signatures are left out, so the proof does not show that every past spend was authorised. The author describes the work as a research preview that has not been externally reviewed.
Read plainly, this is a proof of concept for cheaper bootstrapping rather than a change to how Bitcoin validates. It does show that the space between “download and replay everything” and “trust a snapshot” has room for a third option: trust a proof, verify it cheaply, and keep the trust boundary small enough to describe in a sentence.