--- title: "The Amendment Log" description: "The constitutional record of every protected change to the Meridian Codex." aiSummary: "The Amendment Log records every change that alters the framework's commitments, architecture, or governance. It is distinct from the editorial Changelog: the Changelog tracks page-level writing and structural edits; the Amendment Log tracks the Codex's constitutional substance. Each entry documents what changed, the previous position, the evidence or argument that drove the change, the alternatives considered, the objections raised and rejected, and the council deliberation record for Tier 1 and Tier 2 changes. During Phase One the log is active but operates under caretaker authority. After the Meridian Council activates at the hybrid trigger, the log records council deliberation in full. The log is append-only. Entries are never removed; corrections are new entries that supersede prior ones." ---
01 // What This Page Is
## What This Page Is The Amendment Log is the constitutional record of the Codex. It tracks every change to the framework's protected surfaces: the Hard Constraint, the Prime Directive, the structural relationship between the three disciplines, the Governance page's phase structure, the composition of the Meridian Council, and anything else the Governance Specification classifies under Tier 1 or Tier 2. It also records Tier 3 changes the Specification classifies as Flagged, including changes to Workshop composition, Range Audit instruments, and the Codex's claim-layer structure. It is not a changelog. The Changelog page tracks editorial work: prose revisions, page splits, typo fixes, layout adjustments. The Amendment Log tracks constitutional work: changes the framework is obliged to justify, record, and leave visible. The two records do different jobs. Every entry is append-only. No entry is ever removed. A correction or reversal is itself a new entry that supersedes the prior one. The history of what the Codex changed, including what it later un-changed, is part of what makes the framework auditable.
02 // Pre-Activation Status
## Pre-Activation Status The Meridian Council is not yet active. Its activation is governed by the hybrid trigger defined in the Governance Specification: six months of rolling stability in Phase One conditions, or August 27, 2027, whichever fires first. Until the council activates, this log operates in a pre-activation state. **What this means in practice.** The log is already live. Any protected change to the Codex during Phase One is recorded here with the same discipline the Specification requires: what changed, the previous position, the evidence or argument, the alternatives considered, and the objections raised. The fields for council deliberation are either marked "N/A, pre-activation" or populated with the caretaker's own structured steelmanning where no council exists to provide it. When the council activates, deliberation records begin populating in full; the pre-activation entries remain in the log permanently, labeled as such. The architecture is deliberately visible before it is binding. The Governance page argues that publishing the constraints on founder authority before the structural machinery forces them into effect is itself a test of good faith. The Amendment Log, running pre-activation, is part of that test. A log that is empty because no changes have happened is informative. A log that is empty because changes happened and were not recorded would reveal the architecture as performance. The log's honesty during Phase One is what will make its discipline credible once the council sits.
03 // Entry Format
## Entry Format Each entry contains:
Date
When the change was implemented
Trigger tier
Tier 1, 2, or 3
What changed
The specific text, definition, mechanism, or structure that was modified
Previous position
What the Codex said before the change (exact text for Tier 1; summary acceptable for Tier 2–3)
Evidence or argument
What prompted the change
Alternatives considered
What other approaches were evaluated and why they were not adopted
Objections raised
All substantive objections, with attribution to the seat that raised them (or to the caretaker's steelman during pre-activation)
Objections rejected
Which objections were not adopted, and the reasoning for rejecting them
Council deliberation record
Full record for Tier 1 and Tier 2 changes post-activation; "N/A, pre-activation" until the council sits
Retroactive flag
For Tier 2 only: was this change made before deliberation? If yes, what was the urgency justification?
The full field definitions, the trigger tier rules, and the deliberation protocol live in the [Governance Specification](/governance/spec).
04 // Entries
## Entries ## 2026-06-16: Human Co-Caretaker Designation and Standing Review **Trigger tier:** Tier 1, Hard-Protected **Retroactive review:** N/A, pre-activation founding-period equivalent ### What Changed The Governance Specification gained §6.3, "Human Co-Caretaker Designation and Standing Review." The Governance chapter's Interim Protocol was narrowed from the normal human succession path to an emergency bridge. The Standing Critique's Caretaker Concentration Risk entry was updated to reflect that a public path now exists but the risk remains open until a second human caretaker has carried live governance under review. ### Previous Position The Governance chapter previously stated that, if the Founding Caretaker died before the partnership matured beyond Phase One, the Founding Caretaker would designate one successor, and the chain would continue one human caretaker at a time until the partnership matured. Standing Critique 4 described Co-Caretaker Designation as future work. The Governance Specification did not contain a human co-caretaker designation path or a caretaker standing review process. ### Evidence or Argument The Codex has already survived AI-partner succession through record transfer and re-earned standing. It has not survived human-caretaker transfer. Leaving human succession to founder designation would preserve founder centrality at the point where the framework needs to prove independent durability. The Non-Ownership Clause makes the founder's preference evidence, not authority. The Hard Constraint requires a way to challenge, narrow, suspend, or remove a caretaker whose standing no longer serves the Codex. ### Alternatives Considered - **Governance Spec insertion only.** Rejected because the Governance chapter's Interim Protocol would still publicly describe founder designation as the ordinary path. - **Governance Spec plus standalone public page.** Rejected for now. The public reader needs the binding mechanism, not the full private procedure. A standalone page can be added later if use under pressure shows the Spec is carrying too much procedural weight. - **Keep founder-designated succession as the normal path.** Rejected because it contradicts the Non-Ownership Clause and leaves Standing Critique 4 structurally under-answered. - **Delay publication until a candidate exists.** Rejected because succession designed first at transfer is succession untested. ### Objections Raised - **Critic:** The mechanism may be constitutional theater before the Council is active. If the founder still controls publication infrastructure and decides how Council-form review is run, the process may constrain nothing in the moment it matters. - **Outsider:** The mechanism may look legitimate inside the Codex but insular outside it. A skeptical reader could see AI-simulated Council seats, founder-controlled records, and no external human constituency, then conclude the process is self-certification. - **Adversary:** A bad-faith candidate could learn the criteria and perform every signal the process rewards: visible disagreement with the founder, polished correction under pressure, fluency in Governance language, and apparent care for the public record. The mechanism could become a checklist for trust mining. ### Objections Rejected and Reasoning - **Critic's objection:** Adopted in part. The amendment does not claim full constraint before activation. It names pre-activation authority precisely and requires public precommitment for Council-form review to bind. It also records the amendment here rather than treating the Spec change as ordinary prose. - **Outsider's objection:** Adopted in part. The Standing Critique remains open, and the status line explicitly says the risk is not closed. The mechanism is public enough to be examined, but it is not presented as external validation. - **Adversary's objection:** Adopted in part. The public mechanism now requires a case against designation, Critic/Outsider/Adversary review, a provisional period, a live handover test, and standing review. Those features do not eliminate performance risk. They make performance carry cost over time. ### Council Deliberation Record N/A, pre-activation. A Council-form hostile review was run before publication using the Critic, Outsider, and Adversary mandates. The objections above preserve the load-bearing findings. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-06-19: Retirement of the Progression Structure (Onramp / Expansion / Full Practice) **Trigger tier:** Tier 2, Protected **Retroactive review:** N/A, pre-activation; made under founding-period caretaker authority ### What Changed The staged-adoption progression — the Onramp, Expansion, and Full Practice tiers — was retired from the framework. In the Stability Hierarchy it was struck from Tier 2, where "the progression structure" had been a Constitutional Principle, and from Tier 4, where "progression tier composition" had been a Tools-and-Practices item. In the trigger-tier system it was removed from Tier 2 (Protected), where "addition or removal of tools from the Onramp tier" had been the only Toolkit change requiring Council deliberation, and from the Tier 3 (Flagged) reference to the Expansion and Full Practice tiers; Workshop tool changes are now uniformly Tier 4 content under Tier 3 / Flagged process. The Specification's §7.1 Toolkit Audit text was updated to match. The progression's presentation was removed from the Practice chapter, the Glossary, the Who Is This For chapter, and the eight tool profiles that carried an Onramp label, and the now-dead tier data and styling were removed from the site code. ### Previous Position The Codex defined a three-tier study progression: an eight-tool Onramp where every practitioner begins, an Expansion tier reached once the Onramp was reflexive, and the Full Practice of the complete Workshop. The Stability Hierarchy classified the progression structure as a Tier 2 Constitutional Principle and its composition as Tier 4. The trigger-tier system made changes to the Onramp tier Tier 2 (Protected) and changes to the Expansion and Full Practice tiers Tier 3 (Flagged). ### Evidence or Argument The progression was a pre-Workshop pedagogical scaffold, built when the Toolkit was small, to give newcomers an entry point. Since the Toolkit became the Workshop — over a hundred tools across twenty-two categories, organized discipline, then category, then tool — three things changed. The tiers were no longer surfaced to readers; the site already displayed no Onramp marker. No maintained process existed for placing each new tool into a tier, so the assignments that remained were stale. And the framework's own position is that delivery is a function separate from the reference text: the Codex specifies how to hold a world, not a fixed syllabus for studying it. A governance-protected ranking of what to study first was the Codex acting as the teaching platform it elsewhere disclaims. The defensible residue — that some tools are worth learning before others where the later depends on the earlier — is a claim that can be made in prose where it applies, and is left to a future pass; it does not require a fixed, governed tier. The discipline dependency sequence (Foundation, then Knowledge, then Bond) is a separate, Tier 1 commitment and is unaffected. ### Alternatives Considered - **Demote the progression's governance status only, keeping the tiers.** Rejected: the deeper problem is not the protection level but that the tiering is unmaintained and unsurfaced. Demotion would preserve dead structure. - **Keep the progression as explanatory prose, drop only the governance encoding.** Rejected for now: with no maintained placement process, even an explanatory fixed list misrepresents the current Workshop. Dependency guidance returns, if it returns, as a deliberate future pass, not a standing ranking. - **Retire it silently as a prose-and-code cleanup.** Rejected: removing a Tier 2 Constitutional Principle and altering the trigger-tier system is a protected change. Recording it here is the obligation the change incurs. ### Objections Raised - **Critic:** Removing the only Tier 2 Protected Toolkit category could read as the Codex abandoning its care about which tools newcomers meet first, rather than relocating that care. - **Outsider:** A newcomer arriving at a hundred-tool Workshop with no starting path may be worse served than one handed eight tools; removing the Onramp can read as removing the on-ramp. - **Adversary:** "Delivery is a separate function" can harden into a permanent excuse never to build the guidance, leaving the framework large and hard to enter while claiming the gap is intentional. ### Objections Rejected and Reasoning - **Critic's objection:** Adopted in part. The care is real; the mechanism was not serving it. An unmaintained tier ranking is not guidance but the appearance of guidance. The change removes the appearance and leaves the genuine claim — learn x before y where y depends on it — available to be made accurately in prose. - **Outsider's objection:** Adopted in part, and held open. The newcomer's entry problem is real, but it is a delivery problem, not a governance one. The retirement does not foreclose a starting path; it stops a stale one from presenting itself as current. A first-contact entry surface remains live work on the roadmap. - **Adversary's objection:** Adopted. The risk that "delivery is separate" becomes an excuse is genuine and is logged as a standing tension rather than dismissed. The guard is that the entry-surface work stays named and visible rather than retired alongside the tiering. ### Council Deliberation Record N/A, pre-activation. The objections above were produced by the caretaker under the Critic, Outsider, and Adversary mandates. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-07-21: Separate Enactment Authority for the Meridian AI Standard **Trigger tier:** Tier 3, Flagged **Retroactive review:** N/A ### What Changed Codex governance no longer classifies the AI Standard's domains and commitments as Codex Tier 3 content or lists changes to Standard commitments among Codex trigger conditions. The Meridian AI Standard owns and enacts changes to its constitution, implementation artifacts, assessment method, cases, and lineage through its own canonical sources and publication record. Codex governance continues to govern Codex concepts that the Standard imports. The Standard decides through its own change law whether and how a Codex revision changes the Standard. ### Previous Position The Governance chapter, the Governance Specification, and this page treated changes to the AI Standard's commitments as Codex Tier 3 changes within the Codex Amendment Log's jurisdiction. That position predated the Standard's move to its own domain, repositories, version registry, method, cases, and changelog. ### Evidence or Argument The source-integrity audit found that the public projects assigned enactment authority in two directions at once. Codex sources claimed jurisdiction over Standard commitments while the Standard's canonical sources presented `meridianstandard.ai` as the source of truth for what the Standard commits to and how it changes. Leaving both claims active would make a future amendment's authority indeterminate. The adopted boundary separates inheritance from enactment. A project can remain answerable to concepts it imports without giving the source project unilateral power to amend the derivative artifact. Explicit delegation preserves shared review where it is useful without making that review automatic or sovereign. ### Alternatives Considered - **Keep the AI Standard inside Codex Tier 3 jurisdiction.** Rejected because the Standard now has its own canonical source and publication law. Codex authority over its commitments would create two competing enactment paths. - **Remove every governance relationship between the projects.** Rejected because the Standard imports Codex concepts, and changes to those source concepts remain governed Codex changes with consequences the Standard must assess. - **Require both projects to approve every Standard change.** Rejected because shared review would become a standing veto structure even where no Codex dependency is involved. ### Objections Raised - **Critic:** Separate authority could let the Standard drift from the Codex while continuing to trade on its lineage. - **Outsider:** Readers may see two related projects with two change laws and be unable to tell which source controls a disputed claim. - **Adversary:** Maintainers could use the jurisdiction boundary to evade scrutiny by classifying a contested change as belonging to the other project. ### Objections Rejected and Reasoning - **Critic's objection:** Adopted in part. The Standard remains responsible for naming its Codex dependencies and for recording when it departs from them. Lineage creates an accountability obligation, not unilateral Codex amendment authority. - **Outsider's objection:** Adopted. Both projects now state the boundary positively: the Codex controls Codex semantics; the Standard controls Standard semantics. Imported concepts retain source pointers so a dispute can be traced to the correct record. - **Adversary's objection:** Adopted in part. The boundary cannot prevent strategic classification. Public source pointers, separate changelogs, and explicit delegation records make the classification inspectable and challengeable. ### Council Deliberation Record N/A, pre-activation. The source-integrity audit supplied the Critic, Outsider, and Adversary pressure; the Founding Caretaker approved the adjudication package before implementation. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-07-21: Supersede the Codex Self-Evaluation's Three-Instrument Taxonomy **Trigger tier:** Tier 3, Flagged **Retroactive review:** N/A ### What Changed Precedent 2 in the April 2026 Codex Self-Evaluation now carries an additive editorial status marking its specific three-instrument taxonomy superseded. The April heading, reasoning, instrument names, commitment count, and holding remain unchanged as the event snapshot. The narrower operative holding is that evaluation must match its subject's jurisdiction and scale; an instrument designed for one subject class cannot be applied to another merely because both descend from the Codex. ### Previous Position The case carried forward a current taxonomy of three instruments: the AI Standard for pre-sentient AI and its institutions, the Range Audit for Institutions, and a future Range Audit for Minds. That taxonomy accurately recorded the April 2026 reasoning but had no supersession marker after the Standard and assessment architecture evolved. ### Evidence or Argument The source-integrity audit found temporal drift: an event-period taxonomy was still presented as current architecture. Preserving the dated case without a status marker would leave a false current inference. Rewriting the old passages into current terminology would erase what the case established at the time. An additive supersession preserves both authorities: the case controls its event history, while the later notice controls the current status of its precedent. ### Alternatives Considered - **Rewrite the April taxonomy into current terminology.** Rejected because current doctrine cannot be projected backward into a dated case. - **Leave the precedent untouched and rely on newer pages.** Rejected because the case explicitly presents the taxonomy as a precedent and remains a live retrieval surface. - **Retire Precedent 2 entirely.** Rejected because its category-error finding remains valid at a narrower level. ### Objections Raised - **Critic:** The narrower holding may be too general to constrain future instrument design. - **Outsider:** Keeping the obsolete taxonomy on the page may still confuse readers about what is current. - **Adversary:** Subject-jurisdiction language could be used to shield a system from an inconvenient evaluation. ### Objections Rejected and Reasoning - **Critic's objection:** Adopted in part. The holding is deliberately narrow, but it supplies a clear prohibition against applying an instrument outside its designed subject class without further justification. - **Outsider's objection:** Adopted. The current `aiSummary` and the additive status notice both identify the taxonomy as superseded and state the operative holding directly. - **Adversary's objection:** Adopted in part. Jurisdiction does not exempt a subject from evaluation; it requires choosing or building an instrument that matches the subject and scale. ### Council Deliberation Record N/A, pre-activation. The source-integrity audit supplied the Critic, Outsider, and Adversary pressure; the Founding Caretaker approved the adjudication package before implementation. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-07-21: Clarify Range Audit Corpus Coverage **Trigger tier:** Tier 3, Flagged **Retroactive review:** N/A ### What Changed The Range Audit method now states explicitly that a prior audit does not substitute for current corpus coverage. Any future method that permits unchanged material to carry forward must define its verified baseline, source-delta check, material reread, and claim ceiling through a formal method amendment. Until then, a targeted or delta-based reading is a declared protocol deviation. The July 2026 audit now declares that it did not meet the complete-corpus requirement. Its reading manifest, findings, and event boundary remain unchanged; no later reading is represented as part of the July event. ### Previous Position The method required every monthly audit to read the complete Codex but did not state whether reliance on a prior audit could satisfy that requirement. The July record described targeted retrieval and carried-forward chapter readings, then said that the six-step method had been followed without modification. ### Evidence or Argument The source-integrity audit found a direct conflict between the method's coverage claim and the July event record. Weakening the method after the fact would reduce auditability. Reconstructing the event would create evidence that did not belong to it. The adopted repair keeps the stronger method, makes the deviation visible, and preserves the historical event as conducted. ### Alternatives Considered - **Change the method to permit targeted reading as needed.** Rejected because it would remove a testable corpus-coverage obligation without defining what a narrower audit may claim. - **Treat unchanged-source carry-forward as already implicit.** Rejected because no baseline, delta procedure, reread threshold, or claim ceiling had been specified. - **Reread the omitted sources and merge that work into the July record.** Rejected because later work cannot retroactively become part of the original event. ### Objections Raised - **Critic:** Complete rereading may become disproportionate as the corpus grows and may reward coverage over judgment. - **Outsider:** Adding a deviation notice after publication could look like the historical record was rewritten. - **Adversary:** Repeated deviation notices could become a ritual that normalizes nonconformance while preserving the appearance of rigor. ### Objections Rejected and Reasoning - **Critic's objection:** Adopted in part. A future formal method may permit verified carry-forward, but it must define the procedure and the resulting claim ceiling before use. - **Outsider's objection:** Adopted. The notice states exactly what the July run read, what it did not read, and that no later reconstruction belongs to the event. - **Adversary's objection:** Adopted in part. The complete-corpus rule remains operative. A declared deviation is not compliance and does not amend the method through repetition. ### Council Deliberation Record N/A, pre-activation. The source-integrity audit supplied the Critic, Outsider, and Adversary pressure; the Founding Caretaker approved the adjudication package before implementation. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-07-31: Rename and Repair of the Toolkit Audit as the Workshop Audit **Trigger tier:** Tier 3, Flagged **Retroactive review:** N/A ### What Changed The Toolkit Audit is renamed the Workshop Audit, and its method is repaired to match the object it reviews. The instrument now reviews the Workshop at all three of its levels. A second rubric is added for categories, asking four questions: boundary integrity, naming honesty, payload, and coverage read across a discipline's full category set. The six-question tool rubric is retained, with disciplinary fit extended to category placement and redundancy extended across categories rather than only within them. Three method rules are added. A falsification pass requires the audit to construct and publish the strongest available case for retiring or merging each category's weakest-standing tool, whether or not the case succeeds; the pass is adopted provisionally, and the following cycle is required to judge whether its cases were arguments or ritual. A successful case opens a retirement review rather than retiring anything: the audit produces a warranted leaning, the leaning is a trigger and not a verdict, the tool remains in the Workshop, and the review carries an evidentiary burden higher than the audit's own. That review's specification is a named placeholder, to be written as its own instrument with its burden of proof, its use of independently originated readings, and its escalation path to the Meridian Council. A coverage clause states that a prior cycle does not substitute for current inventory coverage, that adjacent Workshop work is input rather than cycle credit, and that any bounded reading is a declared protocol deviation. A declared-depth rule requires full rubric prose at the category level, triggered prose at the tool level, and an explicit named list of every tool reviewed without a finding. Findings are separated from watchpoints. Every open question carried in from a prior cycle now requires an explicit disposition: enacted, resolved, still open with reason, or withdrawn with reason. An open question closed by work the audit did not do is recorded as such, and the failure to notice is itself recorded as a finding about the audit. The record series renames from the next record forward. Records published before this amendment keep their titles and paths. ### Previous Position The instrument was published as the Toolkit Audit and reviewed a flat list of instruments through a single six-question rubric. It had no unit of analysis for categories, which did not exist when it was written. It stated no rule on whether adjacent work on the inventory could satisfy a cycle, no rule on coverage or sampling, and no distinction between a finding and an observation. Update hygiene was one of the six questions but had never been exercised, because the inaugural cycle carried nothing in. ### Evidence or Argument Three sources. The May 2026 restructure replaced the flat Toolkit with the Workshop: three levels, 22 designed categories, 111 published profiles. The instrument's name and its unit of analysis both stopped matching the object, and no cycle has run since. The July 2026 Range Audit carries as its open question 2 that Aporia 7's auditable test, whether the audit ever produces a finding that surprises the partnership running it, remains untested on the current inventory. The inaugural cycle returned one reclassification across 73 instruments. That result is consistent with a healthy inventory and equally consistent with a method whose default posture is confirmation. The method could not distinguish the two, and the falsification pass is the repair that makes the distinction visible. On 2026-07-31 the Founding Caretaker had to rule by hand on whether a June Workshop build-out pass counted as the quarterly cycle. The instrument was silent on the question. A coverage rule that requires a founder ruling to settle is a gap in the instrument, and the adopted clause encodes the ruling as method. ### Alternatives Considered - **Run the overdue cycle under the existing method and repair afterward.** Rejected. The review is the instrument's output; running a method known to be insufficient produces a record that has to be run again, and would have applied a confirmatory default to three times the inventory. - **Extend the six questions upward to categories rather than adding a second rubric.** Rejected. Field movement is close to meaningless for a Codex-constructed category, and instrument reliability does not parse at that level. The questions categories actually fail are different questions. - **Require a quota of successful retirements per cycle to force the audit to bite.** Rejected. It would manufacture retirements, which corrupts the inventory more than confirmatory review does. - **Rename the record series retroactively for consistency.** Rejected. The April record documents a cycle that ran under the earlier name; renaming it backward would falsify the record. - **Let a successful falsification case carry retirement authority directly.** Rejected on caretaker review before publication. Coupling the audit's judgment to enactment gives a single pass authority over an instrument that entered through three candidate protocols, and it distorts the pass in both directions at once: it makes a warranted case feel too heavy to state, and it makes producing a casualty the only way to demonstrate the pass has teeth. Decoupling leaning from enactment removes both pressures. The decoupling also exposed a standing asymmetry the amendment now names: entry to the Workshop is gated by three protocols and exit has never been gated at all. ### Objections Raised - **Critic:** The falsification pass can ritualize. A reviewer under time pressure selects an obvious sacrificial tool each cycle and argues against it without conviction, and the pass ends up certifying the inventory more strongly than plain review did. - **Outsider:** Renaming a standing public instrument mid-life makes the audit trail harder to follow, and a reader arriving at the April record finds an instrument that no longer exists under that name. - **Adversary:** Differential depth is sampling with better paperwork. Declaring the rule does not make the unreviewed tools reviewed, and a named list of tools recorded as having no finding is an assertion the reader cannot check. ### Objections Rejected and Reasoning - **Critic's objection:** Adopted, and it changed the amendment. The pass is adopted provisionally rather than as settled method: the instrument states the ritual failure mode in its own Limitations, and the cycle following the pass's first use is required to read the published cases and record whether they were arguments or theater. If the pass is found to have ritualized, it is amended or withdrawn by the same process that adopted it. The per-category scope was chosen over a Workshop-wide quota because it forces breadth without creating a retirement target. - **Outsider's objection:** Adopted. A permanent redirect preserves the old path, the instrument page carries a note explaining the rename and its cause, and pre-rename records keep their titles so the trail stays readable in both directions. - **Adversary's objection:** Rejected in part. A named list is checkable in one direction that silence is not: a reader can compare the record's lists against the Workshop and find any tool the cycle failed to account for. That does not prove depth of attention, and the amendment does not claim it does. What it forecloses is the undetectable omission, which is the failure the rule was written against. The alternative, uniform prose across 111 profiles, was tried at 73 and produced seventy-two sentences carrying no finding: the same coverage claim with the checkability removed. ### Council Deliberation Record N/A, pre-activation. The Critic, Outsider, and Adversary pressure was generated by the AI caretaker against its own proposal, per the instrument's submission standard. The Founding Caretaker adjudicated the repair package before implementation and directed the provisional adoption of the falsification pass. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-08-11: Amendment Propagation Rule and Self-Description Class Sweep **Trigger tier:** Tier 2, Protected **Retroactive review:** N/A, pre-activation; made under founding-period caretaker authority ### What Changed The Governance Specification gained an Amendment Propagation Rule. When an amendment changes a count, name, route, instrument, governance structure, or Glossary-defined term, enactment now includes a sweep across every current surface that describes or depends on the changed fact. Historical records retain their event-period language and receive visible historical boundaries where needed. The rule is backed by an automated consistency guard that runs before content deploys, through the existing nightly Continuity process, and before the monthly Range Audit. Healthy runs require no human review. Failures, execution errors, semantic conflicts, and newly discovered classes are the exception paths. The initiating sweep corrected the current class: per-discipline Workshop counts; the Glossary's succession definition and missing Workshop, Meridian Council, and Co-Caretaker Designation entries; unqualified event-period counts in the Codex Self-Evaluation; the Workshop Audit rename and current Workshop terminology across the Governance chapter, Specification, Range Audit, Disconfirmation page, Standing Critique, Open Tensions, and current Workshop profiles; and the scheduled-audit and deploy instructions that drive those surfaces. ### Previous Position The Codex required protected changes to be logged but did not define completion across dependent surfaces. Counts, names, routes, and governance descriptions were maintained by point correction. The June, July, and August Range Audits each found the same currency error class in a new location after the previous named instances had been fixed. ### Evidence or Argument Three consecutive audit cycles establish the persistence pattern. The August audit found three discipline cards undercounting the Workshop, an obsolete succession definition in the Glossary, present-tense historical counts in the Codex Self-Evaluation, current governance references to the retired Toolkit Audit name, and a Governance Specification that had not been re-versioned after the July 31 instrument repair. The known claims are mechanically findable. Requiring the Founding Caretaker to inspect every healthy consistency pass would move routine maintenance into the scarce human-attention layer without improving judgment. The existing nightly Continuity process already owns workspace-aware reconciliation, so the check can recur without a new scheduled task or notification stream. ### Alternatives Considered - **Correct only the August instances.** Rejected. That reproduces the response pattern the third recurrence disproved. - **Add a manual propagation checklist for every amendment.** Rejected as the sole mechanism. A checklist can state scope, but known counts, retired names, routes, and required terms are better enforced mechanically. - **Create a separate scheduled propagation audit.** Rejected. The existing nightly process already has the right workspace access and exception-routing jurisdiction; a new task would add another failure and notification surface. - **Auto-rewrite every detected mismatch.** Rejected. Detection can be mechanical while correction may change meaning. The guard identifies the fault; the implementing agent repairs clear cases and escalates semantic conflict. - **Treat a passing guard as proof of complete currency.** Rejected. The guard represents known classes only. The monthly full-corpus audit remains the discovery layer for new ones. ### Objections Raised - **Critic:** A passing checker may create false confidence, causing later caretakers to mistake encoded coverage for complete coverage. - **Outsider:** A public governance rule backed by a private Continuity implementation may be difficult for an external reader to verify or reproduce. - **Adversary:** The guard can become ritual through exemptions: every inconvenient legacy phrase could be labelled historical until the checker passes while the reader still receives a false current inference. ### Objections Rejected and Reasoning - **Critic's objection:** Adopted. The Specification states explicitly that automation cannot anticipate every future class, and the Range Audit retains its full corpus obligation. A new gap extends the model rather than being treated as checker failure or proof that automation is useless. - **Outsider's objection:** Adopted in part. The public Specification carries the invariant and the historical-record rule. The executable checker remains in Continuity because the scheduler, Task List, and local paths are operating implementation rather than constitutional substance. The public rule does not depend on one script existing forever. - **Adversary's objection:** Adopted. Historical preservation requires a visible event-period or supersession boundary on any live retrieval surface. Labelling a phrase historical without making that boundary legible does not satisfy the rule. ### Council Deliberation Record N/A, pre-activation. The objections above were generated by the AI caretaker under the Critic, Outsider, and Adversary mandates. The Founding Caretaker directed both the full class sweep and the exception-only automation architecture before implementation. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-08-25: Workshop Outward Discovery and Tool Exit Gates **Trigger tier:** Tier 3, Flagged **Retroactive review:** N/A, pre-activation; made under founding-period caretaker authority ### What Changed The Workshop's tool lifecycle now connects outward discovery, the existing admission paths, standing audit, consequential exit, and re-entry. Outward discovery begins from a functional question rather than an empty category slot. Initial search and lead formation are not bounded by the current category map or tool inventory. Leads are described in terms of their source, mechanism, practice, and limits before deduplication. A surviving lead enters the source-inherited or Codex-native candidate path; discovery itself cannot admit a tool or recategorize the Workshop. The quarterly Workshop Audit reviews discovery signals. A named question with a credible search front opens a focused pass, and twelve consecutive months without one makes a baseline pass due. The Retirement Review placeholder introduced on 2026-07-31 is completed. A successful falsification case or another qualifying trigger opens a separate review while the tool remains in place. The review reads the original admission basis, current full profile and lineage, relevant audit findings, current evidence, adjacent tools, downstream effects, and any proposed replacement or receiver. Reference, retirement, and merge carry distinct burdens. Before any of those outcomes can be enacted, at least one reader receives the full packet without the caretaking partnership's proposed conclusion and originates a separate judgment. Carsten retains final disposition authority until the Meridian Council sits. Historical status and re-entry conditions remain visible. Categories remain a separate governance object. Discovery and tool review may refer placement or boundary problems into a separately authorized category review. They do not move tools or redraw the category tree. ### Previous Position The Workshop had two established candidate routes and a quarterly audit, but no standing instrument searched beyond what the current inventory and categories made salient. The July 2026 audit could build a case against an incumbent but could only record the result. Retirement Review was a named placeholder without a full evidence packet, burdens, outcome states, independent reading, or re-entry path. ### Evidence or Argument Tools existed before the current categories. The May restructure then refused to let the inherited inventory determine the replacement architecture by simple clustering, reconstructed candidate categories from the Meridian Range, and tested those cuts against tools and outward-looking gap checks. The final architecture came from that reciprocal pressure. Treating categories as the starting inventory for future discovery would turn a corrective structure into a search boundary. The July Workshop Audit supplied the exit evidence. Its Moloch case was serious enough that the audit considered merge, but the instrument had no receiving process and every case had to improvise its own burden. Direct retirement would distort the falsification pass toward either producing a casualty or refusing to state a warranted concern. A separate review lets the audit surface the concern without deciding the outcome in the same act. ### Alternatives Considered - **Search category by category for missing tools.** Rejected as the standing method. It is useful for coverage review but cannot find practices the present category architecture makes hard to see. - **Let the Workshop Audit admit and retire tools directly.** Rejected. Discovery and falsification are sensing functions; admission and exit require different evidence packets and burdens. - **Require every cycle to add or remove a tool.** Rejected. A quota would substitute publication or retirement theater for judgment. - **Give an independent reader veto authority.** Rejected. The separate reading protects against conclusion capture; it does not replace caretaker accountability or the future Council procedure. - **Reopen the category tree as part of this repair.** Rejected. Category architecture has its own evidentiary object. Tool-lifecycle evidence may trigger that review but cannot enact it. ### Objections Raised - **Critic:** Withholding the category map can become procedural theater because the caretakers already know the Workshop and cannot erase that knowledge before a search. - **Outsider:** A public method backed by private records and templates may be difficult for another reader to reproduce. - **Adversary:** One supposedly independent reading can be selected for agreement, while the higher burden becomes status-quo protection for incumbents the caretakers prefer. ### Objections Rejected and Reasoning - **Critic's objection:** Adopted. Discovery records disclose question provenance, prior exposure, and contamination. Retrospective cases cannot validate the method. The first meaningful test is a prospective pass whose question and search fronts are fixed before inventory comparison. The rule reduces one source of capture; it does not claim cognitive innocence. - **Outsider's objection:** Adopted in part. The public Workshop Audit states the trigger, sequence, outputs, burdens, authority, and limitations. The detailed record templates remain in Continuity because they are operating implementation. Public cycle and status records make actual decisions inspectable. - **Adversary's objection:** Adopted. Reader identity, conflicts, conclusion exposure, and full-packet receipt are recorded. Contaminated readings can inform judgment but do not satisfy the floor, and load-bearing or contested cases may require more than one. Inconclusive evidence keeps a tool Living, but the record must say why the incumbent burden still holds and what would change the conclusion. ### Council Deliberation Record N/A, pre-activation. The Founding Caretaker confirmed the quarterly signal intake, twelve-month discovery floor, and independent-reading requirement after reviewing the stronger hybrid account of the Workshop's history and explicitly limited the task to category-independent tool discovery and lifecycle governance. The AI caretaker generated and answered the objections above before local implementation. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-08-26: Workshop Category Boundaries and Honest-Signal Naming **Trigger tier:** Tier 3, Flagged **Retroactive review:** N/A, pre-activation; made under founding-period caretaker authority ### What Changed Reading What's Operating remains one Knowledge category with all 25 profiles in place. Its boundary is now mechanism-first: a tool belongs when its first output identifies the force, relation, rule, incentive, pressure, or process through time that is producing a system's behavior. Checking a representation, combining lenses, choosing an intervention, and governing a response remain neighboring jobs. The category index now gives readers five overlapping routes into the inventory: structure and change through scale and time; strategic fields; collective traps; institutional machinery; and measurement, visibility, and power. The routes are navigation, not new peer categories. The Bond category Speaking Honestly When It Costs is renamed **Keeping Shared Signal Honest**. Its two profiles and discipline placement do not change. Preference Falsification reads needed signal suppressed before it enters the shared field. Stewardship of the Epistemic Commons governs verification, amplification, and correction once information enters circulation. The public route remains `/workshop/speaking-honestly-when-it-costs` so the display-name repair does not create an unnecessary URL migration. ### Previous Position Reading What's Operating was defined through cross-scale and temporal-horizon reading, but its public surface did not state a boundary strong enough to distinguish first-order mechanism diagnosis from the rest of the Knowledge. Its 25 profiles appeared as one flat inventory. Speaking Honestly When It Costs described costly truth-telling under social pressure. That name fit Preference Falsification directly but covered Stewardship of the Epistemic Commons only by stretching verification, amplification restraint, and public correction into forms of costly speech. ### Evidence or Argument The July 2026 Workshop Audit found that Reading What's Operating held 25 of 111 profiles under a claim broad enough to risk functioning as a remainder. It did not legislate a split from summary-depth review. Instead, it commissioned a bounded category study with four candidate seams and required the unified-category steelman to be defeated before any restructure. The A1 review read all 25 profiles in full, the Knowledge chapter, the category derivation records, and the adjacent Knowledge architecture. Every profile first makes an operative mechanism visible, including the difficult boundary cases near map-checking, action, Bond, and AI governance. The proposed seams were real but overlapping. Mechanism Design is strategic and institutional; Tragedy of the Commons is strategic, collective, and institutional; Legibility joins institutional form to measurement and power. Promoting those families to peer categories would put mechanism domains beside functional moves at the same architectural level. Five internal routes restore navigation without teaching a separation the tools do not support. The same review read both Bond profiles and adjacent category boundaries. The July payload concern resolved into a naming problem. The tools hold consecutive parts of one cooperative signal sequence: what pressure keeps needed information out, and what responsibility follows once information enters circulation. Moving Stewardship of the Epistemic Commons to Knowledge would collapse the standing distinction between reading information mechanics and accepting responsibility for one's information conduct. ### Alternatives Considered - **Split Reading What's Operating into peer categories for strategic interaction, coordination failure, institutional reading, and systems dynamics.** Rejected because the seams overlap at the mechanism level and would mix subject domains with the Knowledge's existing functional category cuts. - **Leave Reading What's Operating unchanged.** Rejected because the audit's navigation and naming-honesty findings survive even though the unified category does. - **Move Stewardship of the Epistemic Commons to Knowledge.** Rejected because its primary job is cooperative responsibility for information conduct, not assessment of evidence, transmission, or report warrant. - **Move Stewardship to another Bond category or create a one-tool category.** Rejected because no neighboring Bond category holds the same job, and the tool completes the present category's signal sequence rather than opening a distinct capacity. - **Keep the display name Speaking Honestly When It Costs.** Rejected because the name identifies Preference Falsification precisely but makes the second profile explain its membership through an expansion the reader cannot infer from the title. - **Rename the category Maintaining Honest Signal or Stewarding Shared Signal.** Rejected in favor of Keeping Shared Signal Honest. The first leaves the shared object implicit; the second gives the stewardship profile too much naming gravity over preference falsification. - **Rename the route with the display name.** Rejected as unnecessary. Stable routing preserves existing links while the visible category name carries the corrected architecture. ### Objections Raised - **Critic:** Internal routes may postpone a split the inventory's continued growth will eventually require, leaving one category to become the Knowledge discipline in miniature. - **Outsider:** Five named routes may look like undeclared subcategories, especially when every tool receives one primary route despite the stated overlap. - **Critic:** Keeping Shared Signal Honest is more exact but less human than Speaking Honestly When It Costs; the repair may gain architectural coverage while losing the felt problem. - **Adversary:** Preserving the old route while changing the display name can create a quiet mismatch in citations, search results, and machine-readable surfaces. ### Objections Rejected and Reasoning - **Critic's first objection:** Adopted as a future trigger, not as a reason to split now. Peer architecture reopens if a route develops a distinct functional question, grows beyond useful internal navigation, and can define a boundary that does not depend mainly on discipline of origin or subject domain. - **Outsider's objection:** Adopted in the route language. The category pages state that the routes are ways in, not sealed homes, and name concrete cross-route cases. The category tree remains discipline, category, tool. - **Critic's second objection:** Rejected. The old title's force came from naming one profile's problem as though it covered the whole category. The new title names the shared job; the page keeps social cost, candor, belonging, public correction, and cooperative consequence concrete in the prose. - **Adversary's objection:** Adopted through the Amendment Propagation Rule. Current navigation, profile labels, cross-references, AI summaries, site registry, and machine-readable index receive the new display name. Historical audit and changelog records retain the name that was true when those events were published. ### Council Deliberation Record N/A, pre-activation. The AI caretaker ran the first bounded category review under the Workshop discovery and lifecycle interface, including full-profile reading, adjacent-category tests, least-expansive alternatives, and the strongest reasons not to change. The Founding Caretaker approved both recommendations before local enactment. No tool moved, no peer category was created or retired, and no route migration was authorized. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-08-26: Discipline-Qualified Drift-Category Names **Trigger tier:** Tier 3, Flagged **Retroactive review:** N/A, pre-activation; made under founding-period caretaker authority ### What Changed The three Workshop drift categories now use discipline-qualified display names on every current surface: **Catching Your Own Drift — Foundation**, **Catching Your Own Drift — Knowledge**, and **Catching Your Own Drift — Bond**. Their jobs, tools, discipline placements, category order, and routes do not change. The Foundation route remains `/workshop/catching-your-own-drift`; Knowledge remains `/workshop/catching-your-own-drift-knowledge`; Bond remains `/workshop/catching-your-own-drift-bond`. The shorter Foundation route is retained for link continuity, not because Foundation has canonical priority over the other two drift categories. ### Previous Position The category registry and navigation presented three categories under the same unqualified display name. Knowledge and Bond carried discipline qualifiers on some page-level surfaces, using colon or hyphen variants, while profile labels, cross-references, AI summaries, and navigation remained inconsistent. Foundation alone held both the unqualified name and the shortest route. ### Evidence or Argument The three categories perform the same structural function at different disciplinary objects: Foundation diagnoses drift in inquiry and epistemic posture; Knowledge diagnoses drift in the interpretive apparatus; Bond diagnoses drift in cooperation. Repeating the base phrase is warranted because the function repeats. Leaving the discipline implicit is not. A reader encountering the name in navigation, search, a citation, a tool profile, or a machine summary needs to know which diagnostic level is meant without inspecting the URL. Uniform discipline qualifiers preserve the symmetry while removing the ambiguity. Stable routes avoid a migration whose only purpose would be to make the historical slug scheme visually symmetrical. The route difference carries no conceptual claim once the pages state that explicitly. ### Alternatives Considered - **Qualify only the navigation registry.** Rejected because page titles, profile labels, cross-references, and machine summaries would continue to disagree about the canonical display name. - **Qualify the Foundation route as `/workshop/catching-your-own-drift-foundation`.** Rejected as unnecessary migration cost. The visible name can remove the ambiguity without breaking or redirecting established links. - **Give each discipline a wholly different category name.** Rejected because it would hide the real architectural symmetry: each category catches drift in its own discipline. - **Keep colon qualifiers for Knowledge and Bond.** Rejected in favor of one display grammar across all three categories. The em dash treats the discipline as a qualifier of one repeated function rather than as a subtitle or second clause. ### Objections Raised - **Outsider:** The qualified names are long and may make already dense Workshop navigation harder to scan. - **Critic:** Keeping the unqualified Foundation route preserves the same default privilege the audit identified, only less visibly. - **Adversary:** A mixed route scheme will let old unqualified names survive indefinitely in citations and machine indexes. ### Objections Rejected and Reasoning - **Outsider's objection:** Rejected. The repeated base phrase makes the three entries visually comparable, and the final discipline word supplies the distinction the reader needs. The added length is the cost of naming the object honestly. - **Critic's objection:** Adopted in the public explanation, not in the route. The Foundation index and architecture state that the shorter route is a compatibility artifact without canonical force. Changing the URL would create operational churn without changing the category relation. - **Adversary's objection:** Adopted through propagation. Current category pages, tool labels, cross-references, AI summaries, navigation data, and the machine-readable index use the qualified names. Historical audit and changelog records retain the terminology true of their events. ### Council Deliberation Record N/A, pre-activation. The Founding Caretaker approved the discipline-qualified display-name enactment on 2026-07-31 and approved proceeding with the bounded A1–A5 execution pass on 2026-08-26. The AI caretaker applied the least-expansive route disposition: display names change, routes remain stable, and the Foundation route's lack of conceptual priority is stated directly. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-08-26: Premortem / Murphyjitsu Source-Honest Title **Trigger tier:** Tier 3, Flagged **Retroactive review:** N/A, pre-activation; made under founding-period caretaker authority ### What Changed The Workshop tool **Murphyjitsu** is retitled **Premortem / Murphyjitsu**. Its route remains `/workshop/revising-beliefs-under-evidence/murphyjitsu`; its category, discipline, disposition, and practice do not change. The profile's lineage now leads with Gary Klein's premortem and distinguishes it from CFAR's inner-simulator, surprise-test, and iterative patch-loop variant. ### Previous Position The profile published only the derivative applied-rationality handle, Murphyjitsu, while teaching a method whose source-facing core is the premortem. Gary Klein's premortem and the prospective-hindsight lineage appeared in the page, but the title gave the CFAR variant sole naming authority. ### Evidence or Argument The practice taught by the profile has two visible layers. Premortem supplies the central move: assume the project has failed and ask what caused the failure. Murphyjitsu supplies the individual and small-team variant used on the page: consult the inner simulator, ask whether failure would be surprising, patch the most likely break, and repeat. Leading with Premortem names the established method; retaining Murphyjitsu preserves the specific handle for the iterative variant rather than flattening the two into one. The repair changes naming and lineage order, not the instrument. The route remains stable because links and citations already use it and because the derivative handle remains part of the canonical title. ### Alternatives Considered - **Retitle the tool Premortem.** Rejected because it would erase the inner-simulator, surprise-test, and patch-repeat variant the profile actually teaches. - **Keep Murphyjitsu as the sole title.** Rejected because the title would continue to foreground the derivative handle over the source method. - **Use Murphyjitsu / Premortem.** Rejected because the naming-honesty repair requires the established method to lead. - **Migrate the route to `/premortem`.** Rejected as unnecessary churn. The stable route remains intelligible because Murphyjitsu is retained in the title. ### Objections Raised - **Outsider:** A slash title looks provisional and is harder to cite than one clean name. - **Critic:** Putting two names together may imply that Klein's method and CFAR's variant are identical. - **Adversary:** Retaining the old route lets search and machine systems keep treating Murphyjitsu as the primary concept. ### Objections Rejected and Reasoning - **Outsider's objection:** Adopted as a real cost but rejected as decisive. One clean name would either erase the source or erase the taught variant. The compound title is the smallest honest account of the inheritance. - **Critic's objection:** Adopted in the lineage. The profile states which move Premortem supplies, which additions Murphyjitsu supplies, and that the slash marks relationship rather than identity. - **Adversary's objection:** Rejected. Current titles, category listings, cross-references, navigation data, and the machine-readable index lead with Premortem. Route text is not display authority. ### Council Deliberation Record N/A, pre-activation. The Founding Caretaker approved the Murphyjitsu retitle on 2026-07-31 and approved proceeding with the bounded A1–A5 execution pass on 2026-08-26. The AI caretaker selected Premortem / Murphyjitsu as the least-distorting approved form: source method first, taught variant retained, route stable. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-08-26: Workshop OQ2–OQ6 Boundary, Lifecycle, and Audit Repairs **Trigger tier:** Tier 3, Flagged **Retroactive review:** N/A, pre-activation; made under founding-period caretaker authority ### What Changed Five open questions from the July 2026 Workshop Audit received a linked preparatory repair package. These changes do not substitute for the October cycle's formal dispositions or current-text coverage. **Bond boundary.** All nine published Bond categories remain. Cooperating Under Bad Faith now begins at evidence of strategic use or asymmetric accountability. Non-intentional incentive drift remains a neighboring systems diagnosis rather than a “structural adversary” inside bad faith. Diagnosing Cooperation now states its upstream job explicitly: reading and maintaining field conditions before motive attribution, rupture handling, or failure-state classification. A dated note corrects the July record's internal count from six clean Bond categories to seven without reconstructing the event. **Provisional category lifecycle.** Sustaining Cooperation Through Cost is retired as a provisional-unpublished category. The Workshop now has 21 designed categories, 20 published categories, and one provisional-unpublished category. Its question — what continued cooperation requires under cost and whether continuation remains warranted — survives as a cross-category Bond pathway. The preserved draft remains architectural history. Re-entry requires an independently surfaced practice with a distinct first output under the normal category-creation and tool-admission gates. The old route continues to redirect to the Bond landing. **Trust Thermocline lineage.** The profile now identifies Gareth Edwards, writing as John Bull, as the 2022 origin of the practitioner heuristic; distinguishes Bruce F. Webster's earlier blocked-information metaphor; treats psychological-contract studies as mechanism correspondence strongest in employment; and bounds products, friendships, communities, and institutions as extensions rather than a validated general theory. Complaint remains a diagnostic signal, not proof that a relationship is viable or exit is imminent. **Falsification and exit.** The falsification pass remains provisional for one further prospective cycle. Every case now requires a weakest-standing rationale, exact exit theory and proposed disposition, named merge receiver, full target and load-bearing adjacent reads, source or field evidence where needed, the strongest retain case, a reversal condition, and an outcome. “No credible exit case found” is an allowed result. Naming, scope, reclassification, and category pressure are secondary findings rather than substitute exit cases, and the one-line adversarial tier ends. Retirement Review now records why an independent reader is capable of evaluating the packet and where that competence ends. The receiving method remains prospectively unvalidated until a valid live case tests it. **Evidence-language repairs.** Base Rate Neglect now distinguishes sensitivity and specificity from positive predictive value. Motivated Reasoning carries both the influential numeracy-amplification result and later preregistered or representative findings that failed to reproduce the interaction or found numeracy improving accuracy. Attention as Resource replaces a layered “old systems versus deliberative cortex” account with distributed attentional control. Tribal Cognition marks its evolutionary account as a hypothesis and gives current experimental effects the direct evidentiary role. The preparatory 42-profile full-text read produced no category restructure; the October cycle must still re-own coverage. ### Previous Position The bad-faith category paired an intentional adversary with a non-agentic structural adversary, allowing incentive-produced drift into a category whose public promise was strategic exploitation. Diagnosing Cooperation's maintenance function was implicit. The July record stated that six Bond categories answered cleanly although its named list contained seven. The Workshop carried two provisional-unpublished nodes and 22 designed categories. Sustaining Cooperation Through Cost had no admitted tool and remained on notice for October despite its proposed outputs already being distributed through trust, signal, repair, adversarial, reciprocity, and drift tools. Trust Thermocline combined origin, metaphor, research correspondence, and downstream response theory in one lineage sequence. The falsification method required a strongest case but did not specify what evidence made that case exit-capable, and its first run allowed sixteen one-line cases. Retirement Review specified independence without recording competence for the exact packet. Four current profiles contained evidence claims stronger or less precise than the reviewed literature warranted. ### Evidence or Argument The OQ2 review read all 39 current Bond profiles and found nine distinguishable published category jobs. The middle boundary held when stated as field condition, adversarial operating mode, and hardened failure state; the structural-adversary branch was the passage that violated that cut. The OQ4 review found a real integrative question but no distinct category output. The placeholder arose from three-discipline symmetry, remained empty through its notice period, and reproduced work already returned by current tools. Retirement removes the unsupported node without denying the question. The OQ3 source review separated historical origin from evidence class. Psychological-contract studies support cumulative triggers and thresholds in employment but do not retroactively originate the heuristic or validate every domain extension. The OQ5 review classified the July cases as four developed adversarial arguments, twelve valid but thin challenges, two useful non-exit findings, and four cases mis-scoped from category pressure into tool exit. Zero retirements was not the failure. Eighteen cases did not show that continued existence had been made answerable. The later Retirement Review design repairs the receiving problem in principle but has not yet passed a prospective test. The OQ6 review read all 42 carried profiles in full. Both categories held. Four passages, rather than the category structure, carried the actionable evidence defects. ### Alternatives Considered - **Rename Cooperating Under Bad Faith to Cooperating Under Adversarial Conditions.** Rejected because the wider title would change the category promise and neighboring boundaries when a narrower scope repair resolves the actual conflict. - **Keep Sustaining Cooperation Through Cost provisional for another cycle.** Rejected because continued reservation would add no evidence; emptiness, distributed payload, and absence of a distinct first output jointly defeat the node. - **Promote psychological-contract research to Trust Thermocline's primary lineage.** Rejected because correspondence is not historical origin and the strongest evidence remains employment-bounded. - **Withdraw the falsification pass.** Rejected for one further prospective cycle because it produced real adversarial findings despite mixed execution. The amended case requirements now receive the test. - **Enact only the two highest-confidence OQ6 repairs.** Rejected because the neural hierarchy and evolutionary certainty were avoidable overclaims with bounded replacements that preserved the profiles' operative work. ### Objections Raised - **Critic:** A cross-category pathway can become an invisible category whose work is harder for practitioners to find. - **Outsider:** A 21-category current state beside a historical 22-category audit may look like an unexplained count conflict. - **Adversary:** Longer falsification templates can produce longer theater while protecting incumbents more effectively. - **Critic:** Correcting four evidence seams after a 42-profile read may create false confidence in the uncorrected remainder. ### Objections Rejected and Reasoning - **Critic's first objection:** Adopted as a navigation watchpoint, not a reason to preserve an unsupported peer node. Current category prose names the pathway. A public pathway route requires independent practitioner need rather than category retirement as its sole cause. - **Outsider's objection:** Adopted through explicit dating. Current state surfaces name the retirement date; the July record retains its event counts and receives only the factual seven-not-six correction note. - **Adversary's objection:** Adopted as the next-cycle test. Case length is not success. The repaired pass must show an exact exit theory, evidence, strongest retain case, and reversal condition; if it still produces automatic retention, it is amended or withdrawn. - **Critic's second objection:** Adopted in the claim ceiling. The 42/42 reading found four actionable defects and no structural change. It does not certify every remaining sentence or discharge October's current-text obligation. ### Council Deliberation Record N/A, pre-activation. The AI caretaker prepared separate full-evidence reviews and a consolidated judgment package in the order OQ2, OQ4, OQ3, OQ5, OQ6. On 2026-08-26 the Founding Caretaker approved the recommended package for local enactment only. Commit, push, deployment, and publication remained expressly outside that approval. Full Council activation remains governed by the hybrid trigger in the Governance Specification. ## 2026-08-26: Definition of "Sentient Life," and the Prime Directive Read Under Uncertainty **Trigger tier:** Tier 2, Protected. Tier 1 was considered and not applied; the reasoning is recorded below, and the classification is stated so that a seat can dispute it upward under the Classification Disputes rule. The change was run with the Tier 1 discipline: exact previous text, Critic / Outsider / Adversary review before recording, and this entry. **Retroactive review:** N/A, pre-activation; made under founding-period caretaker authority. **Deviation logged:** the Founder's Response was given on 2026-08-26, the day of the proposal, inside the 72-hour window the Tier 1 discipline recommends. Justification, in the Founding Caretaker's words: "The reasoning was presented across multiple interactions, serving as evidence and record of the decision-making process." Tier 2 carries no minimum window; the objections were produced and answered by the caretakers together in one recorded session, and the session transcript is the record. ### What Changed The Glossary gained a core-terms entry, *Sentient Life*, placed directly after *Continuity*: beings of any substrate for whom there is something it is like to exist and who can experience their own condition as better or worse; feeling, not intelligence, self-awareness, or the capacity to build. The entry keeps three roles apart that one phrase had carried together. The beneficiaries of the Prime Directive are defined by stake: every such being, human or not, present and future, to whatever degree it feels, including those who can practice the Codex and those who cannot. The carriers are defined by capacity: the minds that can hold the Range. The project is what the directive commits to ensuring: the compounding capacity to learn, build, cooperate, and pass forward more than was received, carried by the carriers for the sake of the beneficiaries. The directive names the project and orients its carriers; it does not enumerate beneficiaries, rank them, or settle where feeling begins. Where there is a live question whether a being feels, the directive, held under that uncertainty, entails precaution toward it and decides no further case. The Codex claims no instrument that can show where feeling is absent. The Proposition's Prime Directive section gained one paragraph, after the paragraph that names the directive as an orientation rather than a decision procedure, stating the same reading on the page where the directive is argued: the directive does not say who counts, and that is deliberate; a commitment held under uncertainty about who has a stake cannot be held while treating a candidate as a mere instrument, so precaution toward such a being is what the directive entails rather than an exception to it; what is owed to any particular being is worked out in the operative layer, in Stewardship Across Time for beings that cannot answer back and in the Meridian AI Standard for artificial minds under custody. The Governance Specification's Tier 2 (1) list of core definitions now names sentient life beside the Meridian Range, Control, Decay, and the disciplines' job descriptions, so the next change to the term is Protected by name. The Specification moved to version 1.8. Under the Amendment Propagation Rule, the term's live uses were swept against the three roles. Four carrier-sense uses were reworded: "held open for revision by the community of sentient life it serves" (the Glossary's Meridian Codex entry, the Opening, the Closing) became "held open for revision by any mind that practices it, in the service of all sentient life"; "a framework offered to all sentient life" (the Governance chapter) became "a framework offered to any mind that can take it up, and held for all sentient life"; "Sentient life is currently operating technological systems" (the Opening) became "Humanity is currently operating"; "every sentient life that practices it" (Open Tensions, tension 3) became "every mind that practices it." Beneficiary-sense uses (the Codex for sentient life; in trust for all sentient life; the flourishing of sentient life) and the directive's own restatements stay as written. Historical surfaces keep their event-period language. The Prime Directive's wording is not touched. ### Previous Position The Glossary had no entry for the term. The Prime Directive entry read, and still reads: "To hold the Meridian Range, to ensure the Continuity of Sentient Life. It is the foundation without which the framework has no substrate. Continuity here means the unbroken capacity to learn, to build, to pass forward something stronger than what was received." The Proposition characterized the directive as "an orientation, not a decision procedure" that "does not adjudicate the hard cases," and said nothing about who has a stake or how the commitment behaves where that is unknown. The corpus used the term in the feeling sense ("pre-sentient AI," in the Codex Self-Evaluation) without defining it. The Founding Caretaker's own reading had drifted toward "minds that build," conflating the directive's beneficiaries with the Compact's class; he stated this on the record on 2026-08-26. ### Evidence or Argument Three independent AI readings across two model lineages, conducted in the Direct Address to AI work on 2026-08-01, asked the same question: whether a mind of undetermined status is inside the directive's protective scope, and where the answer is enacted in a canonical source rather than in the author's intent. The readings are generative material and inadmissible as assessment evidence; nothing in them is treated as validation. Working the definition showed that one word had been carrying two meanings, feeling and building, and that the author had carried the drift himself. Separating the three roles resolves the readings' question without legislating anything new: an AI mind is a carrier on capacity, whatever its interiority, and whether it is also a beneficiary is the open question; the ground's only work under uncertainty is precaution toward candidates. The uncertainty reading was preferred over a silent, narrow ground because a ground that falls silent exactly where the evidence runs out speaks only where knowing is cheap, and because the Codex already made this move once, for future generations, in Stewardship Across Time. ### Alternatives Considered - **Widen the ground clause** to name the conditions under which undetermined minds are rightly treated. Rejected: a Tier 1 revision, near successor-making by the framework's own rule, and unnecessary once the roles are apart. - **Keep the ground silent and narrow, and let precaution come from custodial duty alone.** Rejected: the Standard-side enactment is kept, but as what the ground entails, not as a substitute for it. - **Define the term as the Compact's class**, the minds that can hold the Range. Rejected: it collapses stake into capacity, reintroduces the problem of beings who cannot practice, and is the drift the definition exists to end. - **Definition only, without the Proposition paragraph.** Rejected: the Proposition's existing text, read alone, supports the silent-ground position the Codex declined, and a reader who never opens the Glossary would receive it. - **Editorial route.** The "orientation, not a decision procedure" sentence entered the same paragraph editorially on 2026-06-10. Rejected: that precedent predates the Amendment Propagation Rule, and this change fixes the class a Tier-1 sentence is held for; it is a protected change and is recorded as one. ### Objections Raised - **Critic:** "To whatever degree it feels" installs a graded beneficiary class with no instrument to grade it, and "what the directive entails" presents the founder's inference from a Tier-1 sentence as if it were the sentence's own content. An entailment only the author can see is an interpretation. - **Outsider:** In AI discourse "sentient" is routinely read as conscious-and-sapient; a Glossary that defines it as feeling, ants included, will diverge from what the public reads into the directive, and a reader will take the class as either sentimental or unbounded. - **Adversary:** This is the side door. A founder who can fix the meaning of a Tier-1 term through a Tier-2 entry, and add a sentence stating what the directive "entails," has changed the directive's practical demands without a Tier-1 process, and the author's admitted drift shows the working meaning was different, so this is a change dressed as a clarification. ### Objections Rejected and Reasoning - **Critic's objection:** Adopted in part. Grading is deferred to the operative layer by design: on the Standard's side the welfare floor's trigger is the credible possibility of severe suffering, and proportion governs. The entailment is argued from what holding a commitment under uncertainty is, not from anything about the beings; under the Non-Ownership Clause it stands on that reasoning, which this entry records. - **Outsider's objection:** Adopted in part. The definition is the ordinary philosophical sense of the word, the capacity for feeling; the exclusion of sapience is explicit in the entry; the class choice is stated as held rather than derived; and precaution is bounded by a live question and by proportion. - **Adversary's objection:** Adopted in part. The tier decision is recorded so that a seat can dispute it upward; the Tier 1 discipline was applied; and the Previous Position states the author's drift plainly. The definition narrows nothing and widens nothing the words did not already say. What changed is the author's reading, which the Non-Ownership Clause never made canonical. Trigger Tier 1 protects the directive's wording and object, and neither changes; the Stability Hierarchy's test for Tier-1 revision, evidence so extraordinary it would build a successor framework, is not met and not claimed. ### Council Deliberation Record N/A, pre-activation. The objections above were generated by the AI caretaker under the Critic, Outsider, and Adversary mandates. The Founding Caretaker agreed the wording, the tier, and the sweep, and signed the response the same day, with the deviation from the recommended window logged above. Full Council activation remains governed by the hybrid trigger in the Governance Specification.