Bitcoin Resource Accounting — the research program of Bitcoin
Sahi: a complete accounting system for every long-lived resource Bitcoin consumes
(replicated storage, UTXO state, validation, relay, bandwidth), quantifying how
much of each cost the fee market internalizes. SCCR (storage) is Metric
#1 — the first measured member of the RIR family; the rest are research
hypotheses. Open research on block space economics — the fee market, mempool
dynamics, and the permanent cost of data storage.
SCCR daily readings, 2026-08-02 to 2026-09-17 (N=26,586, T=10 yr, model-spec v2.1.1). Every reading is below 1.0 — fees do not cover the modeled 10-year storage cost. Latest: 0.3196 on 2026-10-08. Data: /data/sccr_history.json.
SCCR is an upper bound. N = 26,586 is the observed reachable population; non-listening/private nodes are unobservable, and counting them (they bear cost too) would lower the ratio. Population scenarios — 50K, 100K, and ~265.9K under a historical 9:1 assumption (not a measured count) — are shown in the sensitivity note.
Nothing matches that filter.
Definitions
Self-contained definitions for the terms used across this library.
SCCR — Storage Cost Coverage Ratio
Transaction fees paid (USD) divided by the modeled 10-year storage cost borne by full nodes (USD) at a node count N. A ratio below 1.0 means fees do not cover the modeled cost of storing the data across the network. Because only listening nodes are observable, the reported ratio is an upper bound: counting non-listening/private nodes would enlarge the denominator and lower it.
N — node count
The number of full nodes bearing the storage cost. Primary source: a measured reachable-node count (validating; a floor) (btcnodes crawl, 2026-09-16); the canonical constant is 26,586, a lower bound. An earlier addrman address sample (≥32,000 gossiped addresses) was never a node count.
T — storage time horizon
The assumed lifetime over which stored data is borne by nodes. T = 10 years.
L_net — modeled network storage cost
The modeled per-block storage cost across the network, in USD. At N=26,586 and T=10 yr it is $4,675.65 per block (model-spec v2.1.1).
Data Confidence & Calibration (2026-09)
The dated, graded evidence under the thesis — historical series, per-block validation, pool hashrate, energy, and reachability transcripts. Grades in data-confidence matrix; live readouts below fetch the frozen endpoints.
2010-2026 per-block SPOT samples (weight/size/tx/fee) cross-checking the SCCR reconstruction legs — and confirming single-block fees are noise vs era means.
What is observable about Bitcoin's node population — reachable nodes (measured), gossip-observed addresses, and what is unobservable — with the Tor/I2P share.
Why a continuous historical UTXO-count series is not obtainable from this environment — sources probed, completion paths recorded, chain-leg cross-check 2.7%.
Pool self-reported hashrate is unreachable from this environment (DNS-blocked / token-gated) — attribution validated internally 7/7, network-total 2.34%.
New Research (2026-09)
The bridge/wrapper audits, the relay-vs-storage correction, and the first-party census instruments added this month.
SCCR 0.4307 is not a solvency verdict: on one fee basis fees fund roughly the first 3 of the 10 modelled years of the node-cost commitment. Includes the horizon trap we nearly published.
Catching a bridge hack that never touched Bitcoin: the detector is the ratio between BTC reserves and token supply, and it fires on both the Symbiosis mint and the Liquid drain.
Base-layer audit of the September 2026 bridge incidents: anchor blocks, the BTC actually issued, and the 1.13-billion-to-1 gap between a claimed token supply and bitcoin.
The population ratio (published, ~1:2) vs the per-block relay split (published nowhere) — and why inbound participation is an upper bound, not a split.
SCCR trend week of Aug 2026 — fees cover less as the fee market cools; the unpriced storage cost is growing. The research team now correlates this live.