Corrections

Data Desk

What we said

The network pages and the ticker showed fee rates such as “1 sat/vB” when the network would clear lower. On 30 September the page read 1 sat/vB for the economy fee, which mempool.space’s precise estimate put at 0.20 sat/vB, and the fee ladder drew every rate as the same full bar.

What the record shows

Fee rates are now read from mempool.space’s precise estimate and printed to two decimals under 10 sat/vB, so the economy fee that read 1 reads 0.20. mempool.space’s whole-number recommendation is no longer shown.

How it happened

The desk read mempool.space’s /v1/fees/recommended endpoint, which reports whole sat/vB and never less than 1, so every rate under 1 sat/vB printed as 1. The review compared the page with that same endpoint, so the check agreed with the error instead of catching it.

Story · published 29 Aug 2026

What we said

pool attribution measures coordination, not miner ownership

What the record shows

Pool attribution shows a coordination surface, the pools that assemble block templates. It does not show that miners coordinate, or who owns them, as the Story's chart caption and its sixth paragraph say.

How it happened

The deck shortened the Story's 'visible coordination surface' to 'measures coordination', and the review did not set the deck against the chart caption, which until 30 September showed only as the image's alt text.

Story · published 29 Aug 2026

What we said

a 204.1 billion increase, or 0.8868 percent

What the record shows

204.1 billion dollars over May's 23,013.9 billion is 0.886855 percent, which is 0.8869 percent to four places. The headline's 0.89 percent and the annualized 5.32 and 5.44 percent were right.

How it happened

The four-place figure was cut instead of rounded when the paragraph was written, and the review checked the headline's two-place figure, not the four-place one.

Story · published 18 Sep 2026

What we said

It is 18.3 percent below that today and has gone 311 days without a new one.

What the record shows

On the day this was published, 18 September 2026, the difficulty had gone 323 days without a new all-time high, not 311. The high was set by the adjustment at height 921,312 at 10:14 UTC on 29 October 2025; the elapsed figure reached 324 days on 19 September. The 18.3 percent shortfall, the height and the date of the high were all correct.

How it happened

The desk measured the gap between the all-time-high adjustment and the most recent adjustment, on 5 September 2026, and printed that interval as elapsed days. Difficulty adjusts roughly every two weeks, so the figure was stale by the age of the current epoch, thirteen days at publication. The error understated the paper's own argument. It was caught the following morning while recomputing every stretch between successive difficulty highs from genesis for a separate mining piece.

Wire · published 7 Sep 2026

What we said

The Wire presented the actors' 03:30 UTC demand for a patch as the latest exchange and closed with "Blockstream has published no patch or post-mortem".

What the record shows

At 09:41 UTC on September 7, forty minutes before publication, Blockstream had signed an on-chain message reading "Bridge nodes are patched, safe to return the funds", which verifies against the key it publishes at blockstream.com/pgp.txt and was re-sent at 11:46 UTC. No public Elements release, advisory or post-mortem existed, which remains true.

How it happened

SG's pre-publication scan filtered on-chain messages by sender and printed only those spent from the actors' or the federation's address; Blockstream's messages arrive from other addresses and were not surfaced before publication.

Wire · published 2 Sep 2026

What we said

The Wire said the new SIGHASH_UNIFIED signature format was something "only the fork's own software supports today".

What the record shows

The Knots 29.4.1 release notes say only that third-party wallets may require upgrades to remain compatible; they do not state that only Knots supports SIGHASH_UNIFIED and name no wallet list either way. An operator page lists two patched third-party wallet builds dated 31 August, described as unaudited. SG had not verified any compatibility list before publication.

How it happened

SG extended the release's compatibility caveat into a claim about all other software without checking wallet support, the same overreach withdrawn from the companion Story on 6 September.

Story · published 3 Sep 2026

What we said

The story used normal majority-chain block production to rebut Knots’ description of a network slowing down, framed ASIC use as a rebuttal to ordinary-computer mining, asserted that compatible wallets meant only the fork’s own software, and repeated an unverified exchange-pricing anecdote.

What the record shows

The release’s slowdown description concerned the minority chain, not the majority chain. The project explicitly supports purpose-built ASICs. SG has not established a complete wallet-compatibility list or independently tested the opt-in per-signature replay protection. The unsupported mining/wallet generalizations and exchange-pricing anecdote have been removed. The corrected story separates these limits from the supporters’ node-cost argument.

How it happened

SG used evidence about one chain to rebut a description of another, generalized beyond what the cited observations established, and included market detail it had not independently verified.

Story · published 3 Sep 2026

What we said

The headline and copy presented nine-times issuance as current and described the measurement as beginning at a difficulty reset. The story also compared the chains’ raw difficulty numbers across different proof-of-work algorithms without establishing a meaningful security comparison.

What the record shows

The retained operator observations cover August 30–September 2: 4,064 height increments in about 77.2 hours, about 8.8 times Bitcoin’s ten-minute target pace. The interval includes two adjustments. This is faster calendar-time issuance, not evidence of a higher eventual subsidy cap; halvings remain height-based. Cross-algorithm difficulty numbers do not establish relative economic attack costs. The title and deck now identify the historical window.

How it happened

SG mislabeled a cumulative observation window, failed to distinguish issuance timing from the height-based subsidy schedule, and juxtaposed incommensurate difficulty figures without the hardware and cost evidence a security comparison requires.

Story · published 3 Sep 2026

What we said

The story called September 2 a Tuesday and described August 8 as enforcement of the BIP110 data restrictions. It also treated the last pre-BLAKE2b height, 961,639, as the end of all shared history with the majority chain.

What the record shows

September 2, 2026 was Wednesday. BIP110 specified mandatory signaling at heights 961,632–963,647, lock-in at 963,648 and full activation at 965,664. The August split preceded the later BLAKE2b change at 961,640; the minority chain’s last SHA-256d height is not the end of all shared history with the majority chain.

How it happened

SG conflated mandatory signaling with full transaction-rule activation, and the later algorithm-change boundary with the earlier chain split. The calendar weekday was not checked separately.

The first public worked example was not ledger-derived and linked Executive Order 6102 to 1 FR 1, an unrelated 1936 Federal Register page. This reviewed row replaces that citation with the order text, removes the unsupported 1974 restoration assertion, preserves the order's explicit exceptions, and replaces the seeded identifier with the deterministic ledger id.