ANCHORS — what each commitment proof is required to prove.
Written 2026-08-30. Append-only: an anchor is added when a proof confirms, never edited.

A timestamp proves a DATE. A gate that accepts any valid proof accepts one made five
minutes ago over tampered content — the digest matches, the walk verifies, and the
guarantee is gone. So the expected anchor is pinned here and the gate requires exactly
it. Re-stamping a rewritten file produces a different block and fails.

Fields: commitment file | state | Bitcoin block | block hash | block time (UTC)

COMMITMENT.txt | anchored | 964701 | 000000000000000000005f7741d3e5e776fc0ec62c3539c27a4f3fd0d4e534b9 | 2026-08-30T06:59:54Z

COMMITMENT-2.txt | anchored | 964709 | 00000000000000000001f1aa06ccc841248b9ac83eb20cafcd7f57efbdce7b29 | 2026-08-30T08:08:56Z
    Confirmed 2026-08-30, eight blocks after COMMITMENT.txt. The pending note that stood
    here is replaced rather than deleted: it recorded, accurately, that this commitment
    rested on four calendar operators for about ninety minutes. It rests on the chain now.

COMMITMENT-3.txt | anchored | 965121 | 000000000000000000003c564bc9e2c5c02c6b3c5cbd4780f7b8d421cc021f1c | 2026-09-02T04:58:21Z
    The CODE commitment. Re-stamped at every release: a change to any listed script means
    a new commitment and a new anchor, and the gate fails in between — that friction is
    deliberate, publishing is the moment the code is pinned. The Phase C release anchored
    in block 964789 (2026-08-30T21:35:13Z); the page then had to record that confirmation,
    which is itself a code change, so this entry returned to pending until the next
    release re-pinned it. The recursion is real and this is where it bottoms out: the FROZEN
    record's anchors never move, and the code anchor is pinned last, at release.
    Re-pinned 2026-09-01 to block 964816 (2026-08-31T02:54:02Z), the release that closed
    the second hardening red-team pass: register threshold pinning, default-deny adverse
    sourcing at every tier, unconditional C1 faith bars, source-level redaction scan, and
    the tenure-slope robust-inference correction. Stamped ~02:00 UTC on 31 August and
    confirmed 54 minutes later; the local proof simply could not be saved for a day
    because a stale .ots.bak blocked the upgrade — the chain had it all along, which is
    the point of the chain. The proof also carries blocks 964815, 964848 and 964865 from
    the other three calendars; the gate pins the first attestation in the proof.
    Returned to pending 2026-09-01 (later): the final pre-publication grade found twelve
    criticisms not yet on the page — several introduced by the slope correction itself — and
    fixing them changed the scripts, so the code was re-stamped. Block 964816 now anchors the
    prior code, not this one. Re-pin when the new proof confirms.
    Re-pinned 2026-09-02 to block 965121 (2026-09-02T04:58:21Z): the release that closed the seven-round final
    pre-publication grade — the privacy list hashed, the slope's mechanism corrected, the
    G2 basis made to follow the spec's enumeration, coding made even-handed across parties,
    and the review process described on the page as what it was. Stamped ~02:10 UTC; the
    earliest of its three confirmations is block 965114 at 02:37 UTC, 27 minutes later, and
    the gate pins the first attestation in the proof, 965121; the proof also carries 965152. Landed by hand from the
    working session after two scheduled tasks fired into empty containers.

COMMITMENT-4.txt | anchored | 964789 | 000000000000000000001eebe1e80e999bbb78cd2fdb715e577d3176f409628f | 2026-08-30T21:35:13Z
    ANALYSIS_SPEC_v3, the binding scoring specification, frozen 2026-08-30 and anchored
    the same day, in the same block as the code release that ships it.