PETER TODD READS THE KNOTS DIFF — POISON-BLOCK CACHE VECTOR FLAGGED, PoW-CHANGE WARNING CALLED FALSE

The forensic read. Peter Todd published a line-by-line code review of the Knots BIP-110 consensus diff over the weekend — Core v29.3 against Knots v29.3.knots20260508, roughly 73,057 lines of diff. This is not a manifesto. This is a Core-adjacent developer publishing a read of the actual code before activation, with the receipts and the file references. The Technologist archetype does this. The Alden and Kratter takes covered the economic and mempool-signaling angles. Todd covered the code.

The cache-vector finding. Todd flags a potential cache-defeating poison-block attack vector in the grandfathering code path — "someone should investigate this further." The mechanism, at plain-language level: the way the fork handles pre-activation blocks with certain properties may allow a crafted block to defeat validation caching in a way that forces expensive re-validation. Not a proven exploit. A flagged surface that needs additional review before activation lands consensus weight.

The false-warning finding. The RDTS_CONSENT warning message in the Knots client claims that "BIP-110 backers are already talking about doing a Proof-of-Work change." Todd disputes the framing as printed to end users — the copy runs ahead of what the actual coordination looks like at this point. Small on its own. Not small when it renders inside a warning banner every affected node operator sees.

The code-hygiene notes. Todd's review also calls out TX_PREMATURE_SPEND error-code misuse (the code is being returned in contexts it was not intended for) and non-standard indentation in the P2SH exemption path. Individually these are code-review comments. Aggregated, they read as a fork that was rushed to a signaling window before the consensus-critical code had received the review pass that a fork of this scope normally gets.

The framework read. Miner signaling for BIP-110 sits at 2.53% as of Sunday morning. The fork's hash-power path is minority at best, non-existent at worst. Todd's line reads accordingly: absent an attack surface being weaponized, the fork does not accumulate enough hash to matter. The Technologist point is not "Knots loses." The Technologist point is: the diff has bugs in it, one of them is potentially exploitable, and the review was done in public before activation. That is the process working.