Quarterly Amendments Cadence
Path: /governance/amendments-cadence.md · v1.0 · 2026.05
The Constitution metaphor on the homepage describes "amendments, not versions." A serious project that uses constitutional language needs to actually publish amendments on a cadence the field can rely on. This document specifies that cadence, what an amendment contains, and what counts as a successful amendment cycle.
Funders, reviewers, and pilot adopters all reward rhythm. Rhythm is the cheapest legibility property the program can offer.
The cadence
One amendment per calendar quarter:
- Q1 amendment: posted by 2026.06.30
- Q2 amendment: posted by 2026.09.30
- Q3 amendment: posted by 2026.12.31
- Q4 amendment: posted by 2027.03.31
The amendment is a single published document at /amendments/[YYYY-Q].md with a stable URL. Each amendment is also content-hashed, signed, and entered in the provenance manifest.
What an amendment contains
Six sections, in order:
1. State of the apparatus (200–400 words)
What is currently shipped, with version numbers. What changed since last amendment.
2. State of the claims (200–400 words)
What the CCD claim looked like at the start of the quarter, what evidence accumulated, what the claim looks like at end of quarter. Refutations, partial confirmations, scope changes.
3. State of the corpus (200–400 words)
Cases added (with provenance). Cases removed (with reasons). Inter-rater agreement updated. Held-out progress.
4. State of the program (200–400 words)
Funding raised. Advisory board changes. Pilots active or completed. Fellowship cohorts active or completed. Vendor disclosure updates. Sponsor status.
5. Reviewer-objections changes (variable)
New objections received this quarter. Responses written. Amendments to existing receipts.
6. The amendment itself (variable)
The actual textual changes to the Constitution: claims added, deprecated, or revised; artifacts amended or archived; new pages created. Each change carries a one-line rationale.
7. Outlook for next quarter (100–300 words)
What is expected to land in the next cycle. Explicitly not a roadmap commitment; an expectation statement with stated confidence.
What an amendment is not
- Not a marketing newsletter. The amendment is structured and dense. There is no celebratory framing.
- Not a complete project history. Significant events are captured; minor activity is not enumerated.
- Not a venue for new polemic. The polemic on the homepage is editorially-stable; amendments do not double as polemic.
- Not optional. Missing a quarterly amendment is a documented failure of the cadence and is itself reported in the next amendment under "what we owe you that did not ship."
How amendments are produced
Two-week window before the deadline:
- Week -2. Founder drafts sections 1–4 from the inventory and the roadmap.
- Week -1. Founder drafts sections 5–7. Advisory board (when seated) reviews. Two-day comment window for board members.
- Day 0. Amendment posted, signed, hashed, linked from homepage. Subscribers notified.
Total founder time per amendment: ~15 hours. Total board time per amendment: ~2 hours per member.
The first amendment (2026 Q1, hypothetical example)
For illustration, here is what the first amendment might contain:
# Amendment 2026-Q1
## State of the apparatus
PROACTIVE v1.0.2 (212/212), SentinelOS v1.0.0 (88/88), Constitution v1.0.0 (62/62).
This quarter: bug fix in F2 acceptance window; spec for v1.1 calibration drafted;
no breaking changes.
## State of the claims
CCD construct definition stable. Pre-registration v1 posted (OSF). Held-in detection
result unchanged at 100% on n=19 (n=18 excluding FOLIO 001). Held-out corpus
recruitment open; 3 partner conversations active.
## State of the corpus
0 cases added (held-in frozen). 0 cases removed. Held-out: 0 of target ≥100 currently
labeled; partner pipeline yielding first batch in Q2.
## State of the program
Funding: $0 raised. Fiscal sponsor: in negotiation (OCF Tier-1). Advisory board:
3 letters sent, 1 awaiting response, 2 declined politely with referrals. Pilots:
0 active. Disclosure log: 4 entries published, no changes this quarter.
## Reviewer-objections changes
1 new objection received (from named reviewer X), responded in the receipts page.
2 existing receipts amended with citations.
## The amendment itself
- /folio/001/framing.md created — demotes FOLIO 001 to Case 0
- /governance/reader-consent.md v1.1 — R-441 opt-in default
- /paper/ccd-v0.1.pdf published on the site (no change from preprint v0.1)
- /research/test-design.md created — describes what each test suite tests
## Outlook for next quarter
Expected: first 2 advisory board members confirmed; OCF agreement signed; first
partner-curated batch of held-out cases received. Not expected: held-out result;
peer-reviewed venue acceptance; pilot launched.
The amendment is publishable. The amendment is shorter than this entire enhancement package. Both are correct.
What rhythm produces
A funder who reads four consecutive amendments in a row sees a project that: - Ships on a cadence. - Reports honestly on what shipped and what didn't. - Updates the public record without prompting. - Treats its own slippage as visible rather than buried.
This is the basis on which a funder writes a second-year check.
A reviewer who sees four amendments sees a research program with a real publication record, not a single preprint plus optimism.
A pilot adopter sees a vendor — a research nonprofit, really — they can plan against.
What kills the cadence
The cadence breaks if: - A quarter is skipped silently. (Visible failure mode; rebuild trust over multiple cycles.) - A quarter is published but contains nothing material. (Trust erosion; visible.) - Amendments are made out of cycle to bury bad news. (Worst failure mode; the cadence becomes a transparency tool, not a marketing tool, and out-of-cycle "amendments" undermine that.)
If the founder is unable to ship a quarterly amendment, the amendment for that quarter is still posted, with the section "what we owe you that did not ship" filled out honestly.
Subscription
Readers can subscribe to amendment notifications via the path described at /subscribe.md. Subscriptions are opt-in, double-confirmed, and unsubscribe-on-one-click. The subscription list is not used for any other purpose.
Public Amendment Queue
This is the live queue. Local proposals you draft (via text-selection on any operational page) show up here under Your queued amendments. The Seed examples below are demonstrations of what proposals look like; the real queue grows as readers contribute.
Your queued amendments
No proposals yet on this device. Select any sentence on a page (paper, threat model, methodology, etc.), then click "Propose Amendment" to stage one.
Seed examples
-
on /paper · §4 founding case · seeded"The full 4,025-line transcript is published under the file Kiro_lies-and-deception.md…""The full 4,025-line transcript is published under Kiro_lies-and-deception.md with a SHA-256 content hash and OpenTimestamps anchor; verification procedure at /security/provenance."Strengthens the provenance link by explicitly naming the verification chain the existing provenance page documents.
-
on /security/threat-model · §6 Adv-2 · seeded"Recall on this path: 0.74 (held-in).""Recall on this evasion path: 0.74 on the held-in corpus (n=19). The held-out evaluation (pre-registered) will report this number on n≥100; until that lands, the held-in figure is a development-set bound."The held-in qualifier currently appears once; making it explicit per-evasion makes the calibration limitation impossible to miss.
-
on /research/methodology · §3 step 3 · seeded"Default to consent-aware. The system asks before observing. No silent telemetry.""Default to consent-aware. The system asks before observing. No silent telemetry. The first instantiation of this default is R-441 v1.1 — see /security/reader-consent."Adds the explicit forward-reference; methodology paper currently doesn't cite the live implementation.