Dataset · v1 · sealed 2026-07-29

The Sleeping Fleet

Every Satoshi-era mining reward, classified one coin at a time.

196,479 P2PK coinbase outputs — every mining reward Bitcoin ever paid to a raw public key — each one classified as asleep or awoken by checking whether that exact coin is still in the unspent set. Plus 157,766 departures dated to the block.

Price

$29

Tables

Rows

546,571

Download

83.6 MB

Formats

Parquet + CSV

Buy this dataset

One-time purchase. Instant download, no account, no subscription.

Everything needed to judge this dataset — manifest, samples, verifier, Bitcoin timestamp — is free below, right now.

Check us before you pay

Free downloads

Most data vendors ask you to trust a description. Here is the actual manifest for the actual product, the actual Bitcoin timestamp, and the verifier that checks them — all of it free, before any money changes hands.

Sample rows

The first 1,000 rows of the main table, and the complete by-year summary. Real rows from the sealed file, not a mock-up.

fleet_outputs sample (154 KB)
fleet_by_year, complete (732 B)

The manifest

Row counts, full schemas, per-column digests, every published figure with the SQL behind it, and the controls the build had to pass.

manifest.json (11 KB)

The proof

The root hash and its OpenTimestamps receipt, committing this exact dataset to the Bitcoin blockchain.

root.hash · root.hash.ots (770 B)

The verifier

Recomputes everything from the files alone. Core check needs no third-party package; install duckdb and it re-derives every figure and re-runs every control.

verify.py (10 KB)

The root of this dataset, committed to Bitcoin:

3b02823f0596e708bdb87efb79fe838eae671c63a24ceb5eb0b7abd8612b2758

Contents

What's in the box

Six tables, delivered as both Parquet (ZSTD) and CSV.
Table	Rows	Parquet	CSV	What it is
fleet_outputs	196,479	13.84 MB	31.22 MB	Every P2PK coinbase output, classified
fleet_keys	192,288	14.28 MB	31.26 MB	The distinct public keys paid
fleet_awakenings	157,766	11.40 MB	23.49 MB	First-spend height per clean key
fleet_awakening_years	18	< 0.01 MB	< 0.01 MB	Awakenings by calendar year
fleet_by_year	17	< 0.01 MB	< 0.01 MB	Mined vs still-asleep, by year
fleet_meta	3	< 0.01 MB	< 0.01 MB	Build method and caveats

Schema

fleet_outputs          height BIGINT · output_n INTEGER · value_sats BIGINT
                       pubkey_hex VARCHAR · year INTEGER · sleeping BOOLEAN

fleet_keys             pubkey_hex VARCHAR · n_outputs BIGINT · funded_sats DOUBLE
                       first_h BIGINT · last_h BIGINT · n_sleeping DOUBLE
                       fully_sleeping BOOLEAN

fleet_awakenings       pubkey_hex VARCHAR · funded_h INTEGER · spend_h INTEGER
                       n_history INTEGER · clean BOOLEAN

fleet_awakening_years  wake_year INTEGER · keys_awoken BIGINT · btc_awoken DOUBLE

fleet_by_year          year INTEGER · outputs BIGINT · btc DOUBLE
                       sleeping_outputs DOUBLE · sleeping_btc DOUBLE
                       pct_btc_sleeping DOUBLE

fleet_meta             k VARCHAR · v VARCHAR

Quality

The controls this build had to pass

The export aborts and produces nothing if any of these fail. They are cases where the answer is already known, so a wrong result proves the instrument broken rather than the finding interesting. All four are recorded in the manifest and re-run by verify.py against your copy:

block_1_is_sleeping — Satoshi's first reward after genesis. Its script shows 2024–25 activity from inbound homage deposits; the coin has never moved. Any pipeline reporting this as spent is classifying at script level, and is wrong.
block_9_is_awoken — spent at height 170: the Satoshi → Hal Finney transaction.
sleeping_is_a_strict_subset — if every output classified as sleeping, the awoken side of the join silently failed. That is the failure mode that would make this dataset worthless while still looking complete.
no_negative_values — sanity floor on the value column.

Honest limits

What this is not

It covers P2PK coinbase outputs only — mining rewards paid to a raw public key. It is not a census of all early Bitcoin.
"Satoshi-era" describes the period, not the owner. We make no claim about who controls any key in this file.
70 reused keys are recorded as deferred rather than dated. We would rather hand you a gap than a guess.
It is a snapshot, at block height 959,975 (unspent set) and 959,976 (chain parse), taken 2026-07-28. Coins wake up; 57 did this year.

Read first

The finding is free

Everything this dataset was built to show is published free and in full, including both tables and the method. Read it → and buy the data only if you want to go further than we did.