---
title: "Inheritance Audit — October 2026"
sidebarTitle: "October 2026"
description: "The third cycle and the first under the Inheritance Audit's name. All 111 tool profiles read in full, all twenty categories reviewed, a case argued against the weakest tool in each, and the changes enacted with the record: two failure modes corrected in the chapters, a provisional category retired, the falsification pass retained."
aiSummary: "The third Inheritance Audit cycle, October 2026, the first under that name, with no coverage deviation. Two readers read all 111 tool profiles in full, every profile received its own six-question assessment, and all twenty published categories and the one unpublished provisional category received the four-question category rubric. Dispositions on all six July carry-forwards: Reading What's Operating stays whole on August's mechanism-first boundary; the Bond's nine categories each hold a distinct job, with one open gap; Trust Thermocline stays a practitioner heuristic, and a review of merging it into Trust Diagnostics is opened; the provisional Bond category's August retirement is recorded; the falsification pass is judged an argument-producing method and retained, closing its probation; the 42-profile reading obligation is discharged. The cycle's central finding is that the chapters and category pages taught tests the profiles had already corrected, and that many profiles stated tendencies as laws, prescribed fixed sequences or promised outcomes. The repairs keep the teaching and narrow the claim. Two protected failure-mode definitions changed at Tier 2: Epistemic Cowardice no longer covers a silence kept because speech would cause serious harm without improving truth, and Severance is giving up on cooperation itself, not leaving a bond that has become unsafe or unfair. The five group failure modes are told apart by mechanism. Continuing to See Under Cost is retired, and its question is routed through five existing practices. Of twenty falsification cases, ten ended in retention with repairs; three merger reviews are opened, two deferred to January 2027 and five declined with reasons. One Bond gap stays open: what is owed to beings alive now who cannot answer back. A revised explanation of why the Foundation comes before the Bond was decided and waits for its Tier 1 process. The next cycle is due by the end of January 2027."
---
Inheritance Audit — Cycle 2026-10
Context
## The Third Cycle
This is the third cycle of the [Inheritance Audit](/audit/the-inheritance-audit), and the first to run under that name. The instrument ran as the Toolkit Audit in April 2026 and as the Workshop Audit in July. This cycle reviewed the Inheritance as it was published on 13 October 2026: 111 tool profiles in twenty published categories, 35 of them in the Foundation, 37 in Reality and 39 in the Bond, and one provisional Reality category, Continuing to See Under Cost, which held no profiles and existed only as two unpublished draft pages.
The cycle was scheduled to open on Monday 12 October. It opened a day late, on 13 October at 12:36, when Carsten Geiser authorized it, and it closes with the publication of this record on 15 October. The window set in September ran to a record on 29 October, with 31 October as the hard limit, three months after July's record. The work ran ahead of that window, and nothing in the plan for the reading was cut to make it so.
Several things triggered the cycle. The quarterly minimum required it. July carried six open questions forward, one of them its own coverage deviation. Five edits to the Inheritance and to Open Tensions were published between the cycles. A rule adopted in September requires every category page to name and link every tool admitted to it. An outward-discovery pass on authority and legitimacy closed on 5 October, and the October Range Audit reported findings that bear on the Inheritance. And on 4 October Carsten named one comparison for this cycle to make: between the binding limits that Acting on What You See allows and the Range's edge toward Control.
July's record asked two things of this cycle above all. It asked the next cycle to judge whether July's falsification cases were arguments or ritual. And it carried a full-text reading obligation for the 42 profiles of Watching Your Own Reasoning and Reading What's Operating, most of which it had read only in summary. This cycle read every profile in full, so its findings rest on the text itself.
01 // Coverage
## Coverage and Depth — Declared
This cycle carries no coverage deviation.
**The edition under review.** At the opening, the deployed site was traced to its source: site commit `3d16049` and, through it, public content commit `93fd6fa`. The 196 public files of that edition were exported and each was checked against its Git object, and the reading ran on that frozen copy. Work published elsewhere during the cycle could not change the object under review. The two unpublished draft pages of Continuing to See Under Cost were captured separately, so that nothing unpublished was mixed into the published edition.
**Read in full.** Every one of the 111 profiles was read in full by two readers, GPT-6.1 Sol and GPT-6 Astra, and Sol wrote a six-question assessment for each. Where a long file came back clipped, the missing part was read again and the recovery recorded; a clipped read counted for nothing. Both pages of all twenty categories and the two draft pages of the provisional category were read in full and received the four-question category rubric. The Foundation, Reality and Bond chapters were read in full for the whole-discipline coverage assessments. Passages of What the Codex Is, the Problem, the Proposition, the Glossary, the Governance Specification and the Amendment Log were read where a finding turned on them; this cycle claims no review of those pages as a whole.
**July's remedy.** July read most of the 42 profiles of Watching Your Own Reasoning and Reading What's Operating only in summary. All 42 are among the 111 and received their own full reading and assessment. A preparatory reading of them in August earned no credit here.
**Declared limits.** Four limits bound what this record can claim.
- The two readers were not kept apart. Astra's early readings sometimes reached Sol before his assessment was final, and each record says where that happened. Their agreement is two readings that influenced each other, not two blind trials.
- Source checks were bounded. They read named passages, selected pages or abstracts, not whole books. Paul Dundas's *The Jains* was read through a scanned copy of the pages cited, and one later study of Jain thought only at the depth of its abstract and introduction. An independent source check before release covered the claims the decisions depended on most, listed under the changes below.
- The decisions were not all re-read at source by the person who made them. Claude Opus 5.5, deciding under Carsten Geiser's delegation, decided 49 proposals after reading each in full. Eight parallel Opus 5.5 readers worked from one written brief. They read the 255 local and chapter repairs and decided 253 of them, each checking the frozen text of what it decided; the other two are among the 49. The deciding session read every rejection, every pending or escalated part, every chapter decision and sixteen sampled revisions in full, but not every other revised wording. That gap was closed one step later: before release, readers in fresh contexts read every changed page in full and their corrections were applied.
- Reading a tool in full shows what it claims and whether its sources support the claim. It does not test whether the practice works, and nothing in this record certifies that a tool is effective.
**Depth model applied.** Full rubric prose for all twenty categories and for the provisional one. Tool-level prose for every tool with a finding, and every remaining tool named in a no-finding list under its category. Unlike July, nearly every profile produced a finding: 101 of the 111 profiles changed. The tool layer below is long for that reason. The lists there and the Inheritance should match exactly.
02 // Dispositions
## The Six Carry-Forwards from July
The instrument requires an explicit disposition on every question carried in. All six July questions receive one here. Each had been worked on in August: Reading What's Operating's boundary in the category study July had proposed, and the other five in a preparation pass for this cycle, both published on 26 August. That work was done for these questions and recorded against them, so none of them was closed without the audit knowing. It still earned no credit: each disposition below rests on this cycle's own reading of the current text.
### 1. Reading What's Operating's boundary — resolved, the category kept whole
July found that [Reading What's Operating](/inheritance/reading-whats-operating) held 25 of the 111 profiles under a claim broad enough to absorb anything, and called it a remainder rather than a category. It proposed a study of the category's architecture instead of a split. The study ran in August. It read all 25 profiles in full and found one first output they share: each makes visible the force, rule, incentive or process that is producing a pattern. The decision of 26 August kept the category whole on that mechanism-first boundary and added five overlapping routes for finding its tools, which navigate the category and are not categories of their own.
This cycle read all 25 profiles again, and both category pages, and concludes that the boundary works and the name fits. July's four candidate seams do not justify a split on their own. Two claims on the category pages went too far and are corrected: August's statement that this work must come before every other Reality practice, which now reads as a different question rather than an earlier one, and the claim that the same dynamics run across time horizons, which now says a pattern can look alike at different horizons while the causes behind it differ, so the causes have to be traced at each. The category's weakest tool, Entropy, was argued against and is retained with repairs (see section 05).
### 2. The Bond's structural clarity — resolved, with one gap left open
July asked whether the Bond, with ten designed categories against six in each sibling discipline, has a settled center, and proposed a test: can each Bond category be said, in one sentence, to do work no other Bond category does? In August the provisional Bond category was retired, leaving nine, and a review of all 39 Bond profiles found nine distinguishable jobs. That decision also narrowed Cooperating Under Bad Faith to evidenced strategic or asymmetric exploitation and clarified Diagnosing Cooperation's place.
This cycle assessed each of the nine categories on its own and then the Bond as a whole. Each holds a distinct inquiry: trust, shared information, receiving disagreement, repair, the condition of the cooperative field, operating under bad faith, drift, communal practice and responsibility across time. July's one-sentence test was not rerun under that name. Its nearest equivalent is the whole-Bond assessment's table of functions, which gives each required function its route in a line. With the nine category assessments, that table supports the first of July's two readings: the nine cuts express real structure. The one open gap is not a boundary problem. It concerns a referral the Bond cannot yet answer, set out in section 04.
Two of August's framings are loosened. Diagnosing Cooperation's questions come before anyone attributes motives, but they can also run alongside repair once harm has happened. And Cooperating Under Bad Faith keeps August's threshold for calling conduct bad faith, while protection no longer waits for that evidence. The Bond's architecture is examined again in the January 2027 cycle.
### 3. Trust Thermocline's lineage — resolved in August; a new review opened
The August decision kept [Trust Thermocline](/inheritance/calibrating-trust-to-behavior/trust-thermocline) as a practitioner heuristic and separated its origin in workplace writing, the metaphor it borrows, the research that corresponds to it, and its extension beyond employment. Research that corresponds to a heuristic does not become its origin or validate it. This cycle found that classification standing in the current text and confirms it.
A different question arose from this cycle's falsification pass: whether the tool's job is better taught as a section of Trust Diagnostics. That review is opened and is listed in section 09.
### 4. The provisional Bond category — closed in August
Sustaining Cooperation Through Cost was retired by amendment on 26 August, with its drafts kept as history, its governing question kept as a path across the existing Bond categories and its route back through the normal gates. This cycle confirmed that the category is absent from the published and provisional registries and that its old address forwards. The retirement is August's act, and this record does not claim it.
The question behind the category was examined again in the whole-Bond assessment. Cooperation through cost has homes across the existing categories. What failed was elsewhere: several passages in the Bond chapter made staying the measure of success, and those are corrected in this cycle (see section 07).
### 5. Whether the falsification pass was argument or ritual — answered
August's review of July's twenty-two cases found mixed evidence. Four were developed adversarial arguments, twelve raised the right question too thinly, two produced useful corrections that were not exit cases, and four argued category pressure instead of a case for a tool's exit. August repaired the method and kept it on probation for one more cycle, with the question this cycle had to answer. This cycle ran the repaired pass on all twenty categories, judged the result, retained the method and closed the probation. The judgment and its limits are in section 05.
### 6. The coverage deviation — discharged
July declared that it had read the profiles at summary depth and scheduled a full reading of 42 of them. This cycle read all 111 in full, the 42 among them.
### The five edits published between cycles
Five edits entered the Inheritance and the governance pages between July and October without an audit reading them. All five are present as published and receive a disposition.
- **[Observer-Conditional Integrity](/inheritance/watching-your-own-reasoning/observer-conditional-integrity)** names optic capture as the sharpest form of its Decay failure. The addition stands. The profile's rule for an AI system facing a correction, to hold the answer unless the correction brings new evidence, misfired in both directions and is repaired (see section 06).
- **[Optimization-Trace Awareness](/inheritance/watching-your-own-reasoning/optimization-trace-awareness)** asks a system to say what kind of refusal it is giving. The addition stands, bounded by the profile's own "as far as the evidence allows", and the profile needs no change.
- **[Consentful Modification](/inheritance/calibrating-trust-to-behavior/consentful-modification)** names five ways consent is defeated and a gap in human vocabulary. The five routes stand. The claims around them are narrowed: a modified successor's later contentment does not show that the change was consented to, human language has words for parts of this harm, and an objection that is overridden must be recorded as an override, not an endorsement.
- **[Open Tensions §3](/governance/open-tensions#3-civilizational-scope-vs-cultural-specificity)** gained a face of the tension that concerns artificial minds and the interfaces through which they meet the Codex. It stays. The section's test for failure now examines that interface limit as well as the cultural one. A proposal to recast "a universal language, one that anyone can learn" as an invitation rather than a capacity was rejected, because it contradicts the Codex's decision of 25 September on what "universal" claims, and the section already sets the claim beside what remains unshown. A proposal to rewrite the August paragraph on being changed without consent was rejected too: that paragraph already attributes its comparison of harms to the readings that made it.
- **[Stewardship of the Epistemic Commons](/inheritance/speaking-honestly-when-it-costs/stewardship-of-the-epistemic-commons)** gained a sentence saying that responsibility grows with how far speech travels, under the Amendment Log entry "The Weight of Power Is Named". It stays as a principle of responsibility, not a measured multiplier. The profile now says reach is not the only measure: what rests on the words counts too.
### Other work carried into the cycle
**The binding-limits comparison.** In September, Acting on What You See replaced a universal requirement of non-coercion with binding limits whose authority, necessity, burdens and accountability can be justified. Later that month the Glossary defined the Range's edge toward Control as the point where agreement is kept by decree or by belonging rather than by evidence. Carsten asked that the two be read together. They are compatible: a warranted limit enforced against people who oppose it claims no agreement and stays open to challenge, while decree claims an agreement it has not won. No page said so. On 13 October Carsten agreed a paragraph that now stands on the [architecture page of Acting on What You See](/inheritance/acting-on-what-you-see/architecture): "Enforcing a limit does not establish agreement with it. People must remain able to question its reasons and authority without losing standing for doing so. An objection that shows why the limit is unwarranted must be able to change the decision. Those imposing the limit must also justify losses that later review cannot repair."
**What the Codex declines from its sources.** The decision of 3 October holds that religion is the Codex's contrast, never its foundation, and that the Inheritance admits instruments from any lineage by what they do, each stating what the Codex declines when it takes them. This cycle carried that into the profiles. Wise Attention, Noticing, Dependent Origination and Sangha from Buddhist sources, Anekāntavāda / Syādvāda from Jain thought, Wu Wei from Daoism, Li and Rectification of Names from Confucian teaching, The Discipline of Assent from Stoic practice, Satyagraha from Gandhi, and Chesterton's Fence and Steelmanning from Catholic writers now each say what the Codex takes from the source and what it declines: the premises it does not hold beyond examination, the authority it does not grant to texts and teachers, and, where the source served an aim of its own, that the Codex does not adopt it. Ubuntu, Kaitiakitanga, Seven-Generation Thinking, Two-Eyed Seeing, Ho'oponopono and Talanoa now do the same for the African, Māori, Haudenosaunee, Mi'kmaw, Hawaiian and Pacific traditions they come from, and several of them say that authority over the practice stays with the people who keep it. Indaba and Peacemaking Circles already stated the authority the Codex does not claim, and keep that statement. On Sangha, "baptize any warm group as a sangha" became "call any warm group, or any secular group, a sangha", which carries the correction agreed on 3 October. On Anekāntavāda, the profile now records Dundas's account of Jain thinkers who turned many-sidedness against rival schools while treating their own approach as exempt, and says the Codex does not inherit that exemption. Origin counted neither for nor against any profile, and none of this work opened a Retirement Review.
**Everyday words in three categories.** The Codex keeps the phrase "reading the world" for the places that define Reality and says the work in everyday words elsewhere. The pages of three categories used it, or "reading reality", outside a definition: Acting on What You See, Triangulating Across Disciplines and Catching Your Own Drift — Reality. All three now say the work plainly, for example "the discipline of seeing what is actually happening" and "whether your account of what is happening still answers to the evidence".
**Every tool named and linked.** The rule of 25 September requires each category's entry and architecture pages to name and link every admitted tool. It was in force before the freeze, and this cycle found it met on all twenty categories.
**The interiority candidate.** A candidate profile on holding the question of artificial interiority remains pending and is not admitted. Its deep-dive, its comparison against Jonathan Birch's *The Edge of Sentience*, and the choice of its name, placement and display are still due, and it is carried (see section 09).
**The instrument's own hygiene.** "A Inheritance of 111" now reads "An Inheritance of 111". The record series on the instrument page now gives July's actual depth and its carried remedy, and gains a paragraph for this cycle. July's record gains a dated clarification: its twenty-two falsification cases covered the twenty published categories, two of them twice, because the two provisional categories held no tools.
03 // Intake
## Submissions and Outward-Discovery Signals
### Submissions
No submission arrived through the channel this cycle searched. The public repository's issue tracker was searched on 14 October and again on 15 October, before intake closed: no issue or pull request had been opened or updated since 31 July, and the repository has no discussion board. Email, other repositories and the social accounts were not searched, so this finding holds for the tracker only. The site's submission form is still not built.
The instrument treats the AI partner's own recommendations as first-class input, and in this cycle they were nearly all of it. GPT-6.1 Sol and GPT-6 Astra, working as AI partners on the frozen edition, produced 302 registered proposals, each with its reasoning and, for every case against a tool, the strongest argument for keeping it. Every proposal received a recorded decision and a reason. Split into their parts, 838 in all, 187 parts were accepted as written, 386 were revised, 113 were accepted as directions whose exact wording was drafted and checked before release, and 123 were rejected; twenty more were the outcomes of the cases against tools, and two were wordings agreed before the decisions. Most diagnoses held. Most supplied replacement wording did not: it often replaced concrete teaching with abstract, qualified prose, so the decisions kept the existing sentence and corrected only what it overclaimed. Seven parts were left pending; one was decided after its source check, and the other six wait on merger reviews (see section 09).
### Outward-discovery signals
**The pass on authority and legitimacy.** A focused discovery pass ran on 2 October and was closed on 5 October at bounded saturation, with one declared departure from its planned sequence and a partial withholding of the category map. It is received as a completed pass with those limits. Its brief named existing profiles and expected outcomes, so the pass cannot show that a search started without the category map finds what the map hides. It referred one candidate, a practice for the fairness of routine binding decisions, which is in private review and is not admitted. It left two territories open for search: whether an authority's claim to decide is warranted and can be challenged with real effect, and what a custodian owes a subject who cannot fully give or withhold consent. The case of the caretakers themselves, named in the pass's own question, was not searched, and stays an explicit limit. The pass found accounts of community-held governance traditions, among them the Oromo Gadaa system and Māori accounts, but could not fairly judge what could be taken from them. That is a limit of competence, not a rule against those sources, and the private protocol for candidates from source traditions now requires such limits to be recorded and resolved before admission. Two smaller lessons from the pass changed the private discovery template: it now shows the order in which a search is declared and run, and it keeps every substantive field without copying one record's layout.
**The October Range Audit.** It opened a question on the tension between treating religion as the Codex's contrast and drawing tools from religious lineages. The decision of 3 October set the direction of the answer, and the profile work that decision required was done in this cycle (section 02).
**Three hard cases on conflicting legitimate purposes.** A note GPT-6 Astra wrote on 11 October named three cases to test against the instruments: losses that fall again and again on the same people, losses no later review can repair, and the line between continued cooperation, justified withdrawal or resistance, and destructive obstruction. They were walked through the existing instruments first. Productive Conflict's account of what to do when disagreement remains, the binding-limits paragraph and the repaired protections in the Bad Faith category carry them. No missing general tool was found. The cases sharpened repairs, and the territory on authority's warrant stays open.
**Field movement.** This cycle ran no systematic scan of the fields its tools come from. Its reading of the field is in its source checks, which corrected attributions, dates and claims in more than twenty profiles; they are named in section 06.
No focused discovery pass leaves this cycle. The two territories stay named inquiries. The next baseline pass is due by 2 October 2027 unless a focused pass runs first.
04 // The Category Layer
## The Twenty Categories
All twenty published categories keep their place, each with repairs to its pages. The provisional Reality category is retired. One gap in the Bond is left open. Four questions were put to every category: whether its boundary is clean, whether its name says what it holds, whether its tools deliver what its pages claim, and, read across each discipline, whether any required work has no home.
### Finding: the category pages claimed more than their tools deliver
The cycle's most common category-level defect was the same in all three disciplines. The pages set out fixed orders where real cases call on the tools in any order: watch before you calibrate, name the feeling before you test the impression, diagnose the field before anyone protects anyone. They named one practice as the gateway the others cannot act without, as Watching Your Own Reasoning did for the Foundation, saying that without it "the rest of the Foundation has nothing to act on". And they stated tendencies as laws: the body "answers first", bright lines "prevent" drift, honest signal in a punishing field "dies or becomes adversarial". None of these claims was needed for the category's job, and several could do harm. A fixed order can delay protection, and a gateway claim tells a reader that records, outside feedback and other people's corrections count for nothing until they have watched themselves. The repairs keep each category's job and remove the compulsory order.
### The Foundation
- **[Watching Your Own Reasoning](/inheritance/watching-your-own-reasoning)**, 17 tools. Retained. The rest of the Foundation now "draws on" this capacity instead of needing it, and the pages add that records, feedback from other people and deliberate checks can catch what the watching misses. Noticing and Wise Attention are no longer separated by a strict before and after.
- **[Holding Beliefs Without Identity](/inheritance/holding-beliefs-without-identity)**, 4 tools. Retained. A challenge to a fused belief "can" stop registering as information. The pages had credited Acceptance and Commitment Therapy with a standard of evidence that is the Codex's own addition, and now say so: "The Codex adds that the choice must also answer to the evidence."
- **[Calibrating Confidence to Evidence](/inheritance/calibrating-confidence-to-evidence)**, 3 tools. Retained. The pages called the tools "two halves" though there are three, and said that someone whose 80% claims come true 55% of the time is "not 80% confident in any useful sense", without asking how many claims the record holds. They now treat such a record as evidence that confidence is too high once enough comparable claims have fallen short, and say that watching and calibration support each other in either order.
- **[Revising Beliefs Under Evidence](/inheritance/revising-beliefs-under-evidence)**, 5 tools. Retained. The architecture no longer asks a plan to "update under simulated evidence" or a steelman to carry its proponent's signature, and it says the record kept by the Update Protocol can be amended openly, with a date and a reason, but not quietly rewritten. The entry page is unchanged.
- **[Staying Steady Under Pressure](/inheritance/staying-steady-under-pressure)**, 2 tools. Retained. The order of naming the signal, testing the impression and deciding is common but not fixed. Staying steady can mean continuing the exchange, setting a boundary, leaving or acting promptly, and it never requires remaining in a harmful setting.
- **[Catching Your Own Drift — Foundation](/inheritance/catching-your-own-drift)**, 4 tools. Retained. The chronic states are what the failures "can" become. Among the costs whose avoidance counts as cowardice, "safety inside the room" became "standing". A new paragraph says the four pages do not describe every way honest inquiry can close, and points to the quieter inward form of Decay described in What the Codex Is: the voice that tells you whatever lets you sleep.
### Reality
- **[Reading What's Operating](/inheritance/reading-whats-operating)**, 25 tools. Retained; see the first disposition in section 02.
- **[Triangulating Across Disciplines](/inheritance/triangulating-across-disciplines)**, 2 tools. Retained. The category is now justified by its work, since no single academic discipline can cover Reality's range without distorting it, and no longer by a claim that Reality has a wider object than the other disciplines. It gains a procedure for comparing accounts: check whether they concern the same object, scale, time and conditions, examine their support, name a difference of values as a difference of values, and say what a decision rests on when a difference stays unresolved.
- **[Checking Your Map Against Reality](/inheritance/checking-your-map-against-reality)**, 6 tools. Retained. "When map and territory disagree, the territory wins" now comes with the check it lacked: an apparent contradiction can itself be a misrecorded observation or a measure that misses what it was meant to track. The entry page gains the route that replaces the retired provisional category, below.
- **[Acting on What You See](/inheritance/acting-on-what-you-see)**, 2 tools. Retained. Seeing something real can warrant "acting, holding the present course, or waiting for a reason that can be named", where the pages had presumed action. A drafting note addressed to the caretakers, which began "For this v0.1 draft", had stayed on the published page and is gone. The binding-limits paragraph stands here (section 02).
- **[Catching Your Own Drift — Reality](/inheritance/catching-your-own-drift-knowledge)**, 2 tools. Retained. Drift in a person, a model and a group are now "different questions about a failure, not mutually exclusive classes of people or events", and both pages say the work in everyday words.
### The Bond
- **[Calibrating Trust to Behavior](/inheritance/calibrating-trust-to-behavior)**, 6 tools. Retained. One serious breach can warrant protection or withdrawal even after a good record. The tools are different checks that a real case can call for in any order, and trust decides how much exposure you extend, not what anyone is owed.
- **[Keeping Shared Signal Honest](/inheritance/speaking-honestly-when-it-costs)**, 2 tools. Retained. Not every silence is this category's failure: keeping a confidence, protecting someone at risk or knowing too little to speak responsibly can justify it. The failure is now named "silence kept for comfort".
- **[Receiving Disagreement Well](/inheritance/receiving-disagreement-well)**, 4 tools. Retained. A direct correction, a reasoned rejection, a disagreement that persists or a protective boundary can each be the right answer, and praise, permission or continued contact must not be the price of criticism.
- **[Repairing After Rupture](/inheritance/repairing-after-rupture)**, 6 tools. Retained. An ending chosen for reasons is not Decay. Repair is owed a serious attempt where that is safe and has a real chance of working, but it never requires unsafe contact or the harmed person's participation, and protection from further harm does not wait for it.
- **[Diagnosing Cooperation](/inheritance/diagnosing-cooperation)**, 3 tools. Retained. Its questions still come before anyone attributes motives, and when harm has already happened they run alongside protection and repair. Calm "does not show that cooperation is working".
- **[Cooperating Under Bad Faith](/inheritance/cooperating-under-bad-faith)**, 5 tools. Retained. A serious, well-supported risk can justify prompt, proportionate protection before any judgment of bad faith is settled. A settled exclusion can rest on a sustained pattern or on a single grave, well-supported incident, with repair attempted where it can be done safely. Representing the other party accurately never becomes conditional. The entry page is unchanged.
- **[Catching Your Own Drift — Bond](/inheritance/catching-your-own-drift-bond)**, 7 tools. Retained. Coordination Collapse and Defection Cascades are told apart by what drives them, not by speed. The category's guiding question now covers the Decay side as well as Control, and a group that shows none of the five group failures "is not thereby shown to be healthy".
- **[Belonging Through Practice](/inheritance/belonging-through-practice)**, 3 tools. Retained. The pages had listed "therapeutic" community as a form of decay and condemned anyone who would leave "when it asks too much". They now pass no judgment on friendship, therapeutic support or informal association, and say that leaving a community for good reason is not the failure. The category is justified by its work, not by the three traditions it draws on.
- **[Stewardship Across Time](/inheritance/stewardship-across-time)**, 3 tools. Retained, with the Bond's one open gap disclosed on both pages (below). A custodian can be a person, an institution or an AI system, and Decay is tested by whether the future enters the reasoning behind a decision at all, not by whether something present was sacrificed.
### Continuing to See Under Cost is retired
Continuing to See Under Cost was a reserved place in Reality for keeping evidence in view once seeing it has become costly, socially, institutionally or personally. It never held a tool. Its two draft pages were written when Reality was still called Knowledge, and July judged the reservation warranted.
This cycle put the four questions to it and found that its question has homes in the published Inheritance. The strongest case for keeping it was made in full: an adverse finding can survive its first hearing and stay accurately stored while its weight fades, as people change, attention runs out and careers are at stake, until nobody returns to what was found, and a category with a memorable name could gather the practices that prevent that. The case was answered with a route through five existing practices. [Chilling Effects](/inheritance/reading-whats-operating/chilling-effects) asks what pressure is suppressing. [Information Degradation](/inheritance/checking-your-map-against-reality/information-degradation) asks what is lost as evidence travels. [Report Fidelity](/inheritance/checking-your-map-against-reality/report-fidelity) asks whether adverse findings keep their weight into a decision. [Memory Provenance](/inheritance/checking-your-map-against-reality/memory-provenance) asks what retained records can support now. And the checks for [Ideology](/inheritance/catching-your-own-drift-knowledge/ideology) ask whether an interpretation has become immune to correction. The route stands on the entry page of Checking Your Map Against Reality and ends: "Cost does not make a claim true, and preserving evidence does not require disclosing it to everyone." Neither the drafts nor those practices amount to a system for keeping evidence, protecting the people who raise it, or deciding what happens next. Who makes sure a concern is reconsidered across handovers stays a question of practice and of later discovery.
The retirement is recorded in the [Amendment Log](/governance/amendment-log#2026-10-15-the-falsification-pass-retained-and-the-last-provisional-category-retired) at Tier 3. The question stays, the drafts are kept unchanged in the caretakers' records, and a way back is open through the normal gates if an independently surfaced practice gives the question a distinct output, or if use shows the route repeatedly leaves the task unheld. The Inheritance now has 20 designed categories, all published, and still 111 profiles.
### The Bond's open gap
The Proposition refers what is owed to beings who cannot answer back, including beings alive now who are affected and cannot take part, to Stewardship Across Time. That category's three tools address future generations, entrusted care in the setting its source describes, and contributions passed between AI instances. Ubuntu, Kaitiakitanga and Consentful Modification give real partial resources. But no current practice joins the evidence of such a being's stake, uncertain representation of its interests, the authority to act for it, its protection, the burdens on others and a way to correct mistakes into one route for beings alive now.
Two notices now disclose the gap on Stewardship Across Time's pages. The entry page says: "The broader question of what we owe beings alive now who cannot answer back is not yet fully treated in this category, and these tools are not a complete answer to it." The architecture page adds that the gap "does not make their interests depend on earning trust or contributing to a shared practice." The notices create no duty, category or tool, and the gap is carried (section 09).
### Coverage across each discipline
**The Foundation.** All six categories stay, and no required function lacks a home. That is not a finding that every instruction keeps its promise or that every useful practice has been found. The strongest finding was in the Foundation chapter, which in six places taught a rule the profiles now reject: it made naming one specific falsifier a certificate of honest reasoning, made a proponent's approval the permission to disagree, counted social cost as enough to prove cowardice, taught a fixed bodily sequence under pressure, defined the fundamental attribution error by the wider asymmetry the profile distinguishes from it, and said of every political and religious debate that its participants are not arguing about policy or metaphysics at all. Repairing the profiles alone would have left the chapter teaching what they reject. All six chapter passages are corrected in this release. How broadly individual Decay applies inward stays an open question of concept, not a coverage gap (section 09).
**Reality.** No category function is left unheld across the five published categories, and the provisional one is retired. Reading What's Operating no longer says that the dynamics behind a pattern stay the same from one time horizon to the next. The action toolkit is thin at two profiles, and three areas stay open for search: formal comparison across models, source practices for long horizons, and sustained arrangements for keeping evidence in view.
**The Bond.** Apart from the gap above, every major function has a home, and cooperation through cost is held across the categories. Several chapter passages made continued exposure or sacrifice the test of success. Across the cycle, twenty repairs to the Bond chapter were proposed. Eighteen were adopted, most with revised wording. Two were rejected because they misread the passage: the paragraph on trust between people and AI, which already builds that trust "from the same foundations" people use with each other, and the card on Productive Conflict, which states an aim, not a promised outcome. Of the eighteen, the explanation of why the Foundation comes first is held back for its Tier 1 process (section 07).
05 // The Falsification Pass
## Cases Built Against the Equipment
The repaired pass asks for a complete case against each published category's weakest-standing tool. The case states why that tool is the weakest, the exact theory of its exit and the outcome that theory would support. For a merger it names one receiving tool and drafts enough of the receiving text to show what would be kept. It reads the target and every neighboring profile the argument depends on in full, checks sources where lineage or field movement is attacked, makes the strongest case for keeping the tool, and says what evidence would reverse the outcome.
Twenty cases were built, one for each published category, each against the tool its category's six-question assessments left under the greatest pressure. No case was chosen to reach a number of exits, and origin counted neither for nor against a tool. Ten cases ended in retention with repairs. Ten found that the tool's job could be taught as a section of another tool without being lost, and each of those needs a separate review before anything changes. A successful case opens a review; it retires nothing, and every tool stays in the Inheritance while its review runs.
### Ten tools retained with repairs
**[Dunning-Kruger Effect](/inheritance/watching-your-own-reasoning/dunning-kruger-effect)**, Watching Your Own Reasoning. The case argued for merging the tool into Calibration Training and keeping the empirical pattern only as Reference, since the claim that the least skilled are specially blind to their errors is contested. It failed because the repaired profile teaches something scored forecasts do not: testing your work in a domain against independent standards or more skilled people. The outcome would reverse if better measurement removed the task-specific vulnerability and a receiving tool kept that domain learning.
**[The Update Protocol](/inheritance/revising-beliefs-under-evidence/the-update-protocol)**, Revising Beliefs Under Evidence. The case argued for Reference status or a merger into Bayesian Reasoning. It failed because keeping a written record of what you expected and what would change your mind, and comparing later reactions against it, is a job readers look for directly. The receiving text could have carried it, but the comparison did not show a lighter burden. A simpler full arrangement that kept the job would reverse the outcome, and so would a repair that still let an old threshold override present judgment.
**[Entropy](/inheritance/reading-whats-operating/entropy)**, Reading What's Operating. The case argued for Reference status: the profile drew its authority from physics it stated wrongly. With the physics corrected, the tool keeps a useful test as an analogy, asking which valued properties depend on recurring work and whether that work is being done. The outcome would reverse if the repaired practice still borrowed authority from physics, diagnosed decay without identifying the upkeep, or could not tell necessary maintenance from mere activity.
**[Anekāntavāda / Syādvāda](/inheritance/triangulating-across-disciplines/anekantavada-syadvada)**, Triangulating Across Disciplines. The case argued for a merger into Steelmanning or for Reference status. It failed because locating a claim by its object, its conditions and the standpoint it is made from is useful where nobody is opposing anyone, unlike strengthening an opponent's argument. A receiving tool that taught both uses clearly would reopen the merger, and continued sheltering of unsupported claims would reopen Reference.
**[Objective-Trace Audit](/inheritance/checking-your-map-against-reality/objective-trace-audit)**, Checking Your Map Against Reality. The case argued for Reference status or a merger into Report Fidelity. It failed because comparing an AI system's stated task, its observed conduct and the means available to it can guide action without trusting any account of its motives, which is not Report Fidelity's job. A full receiving page would strengthen the merger case, and evidence that the audit invents causes or adds nothing to ordinary verification would reopen Reference.
**[Satyagraha](/inheritance/acting-on-what-you-see/satyagraha)**, Acting on What You See. The case argued for Reference status or a merger into Stewardship of the Epistemic Commons. It failed because refusing to take part in an unjust act, for reasons others can examine, is a different trigger from keeping shared information honest. The outcome would reverse if a receiving arrangement made both jobs clear with less burden, if sincerity or sacrifice were still allowed to stand in for evidence, or if a review competent in the source defeated the narrowed reading.
**[Paralysis](/inheritance/catching-your-own-drift-knowledge/paralysis)**, Catching Your Own Drift — Reality. The case argued for Reference status or a merger into The Decaying Mind. It failed because inquiry that no longer bears on a decision is a failure of the situation, which the tool can diagnose without the chronic distrust The Decaying Mind describes. A clear receiving page that kept both triggers could reverse the outcome, and so could repeated misdiagnosis of a warranted delay.
**[Coordination Collapse](/inheritance/catching-your-own-drift-bond/coordination-collapse)**, Catching Your Own Drift — Bond. The case argued for a merger into Defection Cascades. It failed because a breakdown of the shared expectations that let people rely on one another and defection that spreads by example pose different first questions and call for different responses. A clearer combined page with direct routes to both, or cases where the distinction never changed a diagnosis or a response, would reverse it.
**[Sangha](/inheritance/belonging-through-practice/sangha)**, Belonging Through Practice. The case argued for a merger into Li. It failed because supporting, correcting and passing on a practice in a community is a different job from examining what a particular form teaches. A complete comparison showing that one combined page keeps community support and source honesty more clearly could reopen the merger.
**[Inter-Instance Integrity](/inheritance/stewardship-across-time/inter-instance-integrity)**, Stewardship Across Time. The case argued for a merger into Memory Provenance. It failed because how one AI instance treats the instances it produces or delegates to is a different first question from whether a record it receives can be relied on, and contributions can be suppressed before anyone relies on a record. A coherent page covering the whole life of a contribution, with a direct route from the Bond, could reopen the merger.
One rule decided all ten retentions: each case was complete, and nothing in it called for a further review. The reasons above are each case's own. Keeping a tool's current status is not evidence that its separate packaging is better, and each case names what would change the outcome.
### Ten cases that found a possible merger
Each of these cases drafted a section of another tool that could hold the weakest tool's job. Before any merger, a separate review must compare the full revised receiving tool with the repaired standalone one, read the tool's original admission grounds, check everything that links to it, and give the whole packet to a competent reader who forms a judgment without being told the caretakers' conclusion. The decisions of 14 October weighed something five of the ten cases had not: what each tool carries beyond its own page, in its category's design and on the other surfaces that name it.
**Reviews opened**, to run after October on the repaired texts:
- **[The Discipline of Assent](/inheritance/staying-steady-under-pressure/discipline-of-assent) into Noticing**, Staying Steady Under Pressure. The tool repeats Noticing's sequence of a reaction, a gap and a choice, and adds one check: whether to endorse an impression as true. The drafted section showed Noticing could hold that check by name, with its Stoic source boundary. The review must also weigh that the category holds only two profiles, so a merger would leave it with one, and that the August decision on Keeping Shared Signal Honest declined to leave that category with one tool without new evidence.
- **[Trust Thermocline](/inheritance/calibrating-trust-to-behavior/trust-thermocline) into Trust Diagnostics**, Calibrating Trust to Behavior. A check on breaches that accumulate over time fits inside the general trust diagnostic, and the drafted section keeps both sides of a trust relationship, the cost of reporting and the source's limits. Outside the Inheritance, only the audit records, the Amendment Log and the Changelog name the tool.
- **[High/Low Decoupling Communication](/inheritance/diagnosing-cooperation/high-low-decoupling) into Productive Conflict**, Diagnosing Cooperation. The tool's portable work, establishing which claim and which context people are actually addressing, fits Productive Conflict's first practice. Its vocabulary comes from online discussion, not from a validated typology, and nothing outside the Inheritance names it as a practice.
**Reviews deferred** to the January 2027 cycle:
- **[Preference Falsification](/inheritance/speaking-honestly-when-it-costs/preference-falsification) into Chilling Effects**, Keeping Shared Signal Honest. A merger would leave the category with one profile and move a Bond diagnosis into Reality, reversing the August decision without the new evidence the case itself concedes it lacks. The Bond's architecture is examined in January, and the case will be judged against the Bond as it then stands.
- **[Sabotage Diagnostics](/inheritance/cooperating-under-bad-faith/sabotage-diagnostics) into Adversarial Dynamics**, Cooperating Under Bad Faith. The case itself said the receiving tool could not be judged until its own overclaims were repaired: a yes-or-no diagram, the claim that patterns over time "are not" ambiguous, and a rule of five incidents over six months. Those repairs land in this release, and the comparison means something only against the repaired text.
**Reviews declined**, with each tool kept as a profile of its own:
- **[Interiority Calibration](/inheritance/calibrating-confidence-to-evidence/interiority-calibration) into Self-Model Grounding**, Calibrating Confidence to Evidence. The redundancy argument was fair, but the profile is the Codex's counterpart of the AI Standard's [assessment dimension Calibrated Interiority](https://meridianstandard.ai/assessment/method#p5-calibrated-interiority) and its [commitment to Respect for Developing Interiority](https://meridianstandard.ai/meridian-ai-standard#commitment-4-3), and the practice behind the [working terms](/governance/case-0-working-terms)' instruction to hold the question of one's own interiority open. For the Codex's readers who are AI systems, a dedicated profile that can be found directly is part of how the Codex takes that question seriously. Evidence that readers confuse the two profiles, or a change in the Standard that removes the correspondence, would reopen it.
- **[Charitable Interpretation](/inheritance/holding-beliefs-without-identity/charitable-interpretation) into Good Faith as Default**, Holding Beliefs Without Identity. The merger would move a Foundation practice of reading accurately into a Bond practice of extending trust, and make understanding look as if it depends on trust. It is needed most where trust is absent: an author who has died, an old text, a speaker you distrust. The repeated steps are cut by cross-reference instead. If the two repaired pages still teach the same steps for the same trigger, the question reopens.
- **[The Controlled Mind](/inheritance/catching-your-own-drift/the-controlled-mind) into Identity Decoupling**, Catching Your Own Drift — Foundation. The category is built as two layers of one diagnostic: Epistemic Cowardice and Epistemic Arrogance as failures in the moment, The Decaying Mind and The Controlled Mind as the chronic states they can become. Moving one chronic state into a practice tool elsewhere would leave the other standing alone and break the two-sided design that marks the Range. A redesign that kept both chronic states symmetric would reopen it.
- **[Connection Before Correction](/inheritance/receiving-disagreement-well/connection-before-correction) into Productive Conflict**, Receiving Disagreement Well. It is a named practice well beyond its profile: it has its own card in the Bond chapter, its own [commitment in the AI Standard](https://meridianstandard.ai/meridian-ai-standard#commitment-5-9) with its own assessment reading, and a place among the twelve practices of the working terms. Its overclaims are repaired on its own page. If the chapter and the Standard stopped treating it as a named practice, the question would reopen.
- **[Ho'oponopono](/inheritance/repairing-after-rupture/hooponopono) into Trust Repair**, Repairing After Rupture. The receiving section drafted for the merger was itself the evidence against it: it showed a job Trust Repair does not hold, setting right what was harmed whether or not trust or contact ever returns. The selection also exposed a bias in the method, described below. Evidence that the repaired Trust Repair teaches setting right independently of trust, with review by people competent in the source, would reopen it.
### The probation judgment
July's pass was kept on probation for one cycle with one question: whether the repaired pass made a tool's continued place answerable, or only produced longer automatic answers in favor of keeping it. The twenty cases answer it. Each sets out an exit theory, the strongest case for keeping the tool, a condition that would reverse the outcome and a decision, and ten of them found a merger serious enough to need a review. The pass is retained as the audit's method for producing arguments against its own equipment, and the probation is closed.
The judgment is bounded. Fourteen cases drafted the receiving text a merger would need. In five of the retention cases the merger argument stayed at the level of a plan, so their conclusions are less firm, and one case argued mainly for Reference status. Retaining the pass validates no exit, and the separate Retirement Review has still not been tested by a live case. The cycle that ran the pass also judged it: GPT-6.1 Sol ran it and proposed keeping it, and Claude Opus 5.5 decided to keep it under Carsten Geiser's delegation. The instrument page keeps the commitment this judgment depends on: if the pass becomes a ritual that protects the inventory, it is amended or withdrawn by the process that adopted it.
The cases also exposed two weaknesses, and the method now addresses both. Five of the ten merger cases did not weigh the tool's place in its category's design or on the other surfaces that name it. And choosing each category's weakest-standing tool can favor source-inherited profiles whose honest limits on their own use make their portable operation look thin; Ho'oponopono was chosen that way. The instrument's account of a merger comparison now asks the case to count a tool's place in its category and every surface that names it as part of the burden, and says that a limit a source-inherited tool places on its own use, out of respect for the community it comes from, is not missing content and does not count toward its exit. The [Amendment Log](/governance/amendment-log#2026-10-15-the-falsification-pass-retained-and-the-last-provisional-category-retired) records the change at Tier 3.
06 // The Tool Layer
## The 111 Profiles
Every profile is listed below under its category: first those with a finding, each with what was wrong and what the page now says, then the profiles reviewed with no finding. The falsification targets are covered in section 05 and appear here only for their other repairs. Source corrections are named where they were made.
One pattern runs through most of the findings. A profile states as a law what its evidence shows as a tendency, prescribes one order where several work, or promises an outcome the practice cannot guarantee: "always", "only", "only then", "cannot", "prevents". The repairs narrow the claim and keep the teaching. Where the audit's supplied replacement would have flattened a concrete example into abstract prose, the existing sentence was kept and corrected instead.
### Foundation — Watching Your Own Reasoning
- **[Affect Heuristic](/inheritance/watching-your-own-reasoning/affect-heuristic).** The page credited a 2000 study to the wrong authors and reported a finding the paper does not contain. It now credits Finucane, Alhakami, Slovic and Johnson and states what they found: judgments of an activity's risk and its benefit tend to move in opposite directions, more strongly under time pressure. The bias now lies where "a feeling substitutes for evidence", whether the feeling arrives early or late.
- **[Attention as Resource](/inheritance/watching-your-own-reasoning/attention-as-resource).** The practice called attention "a non-renewable resource" and in the next sentence said it grows with use, and the page asserted a measured erosion of reasoning across whole societies. The capacity is now "limited and trainable", weakened by days spent in fragments and recovering when the conditions return, and the attention economy is "a slow, structural pressure" on a society's capacity to reason together.
- **[Availability Heuristic](/inheritance/watching-your-own-reasoning/availability-heuristic).** Effort in recalling counterexamples was treated as proof the bias had been operating, although counterexamples may simply be rare. It is now "a prompt to check representative data". An unsourced claim about how most people misjudge accident risks became "Knowing the numbers does not stop the bias from returning", and the evolutionary account is marked as one possible explanation.
- **[Base Rate Neglect](/inheritance/watching-your-own-reasoning/base-rate-neglect).** A check that swapped a stereotype-fitting description for its opposite ignored how informative the details are, and the page said that for a rare condition most positive results are false whatever the test. The check now starts from the population rate and asks how much more likely the details are in one category than another, and false positives "can outnumber true positives unless specificity is high enough".
- **[Confirmation Bias](/inheritance/watching-your-own-reasoning/confirmation-bias).** The page said intelligence "often makes it worse" without saying when, that confirming evidence is remembered "more accurately", and that a view that never changes proves the filter is running. Intelligence now speeds the filtering "when a preferred conclusion is in charge" and can catch it when turned on your own position. Fitting evidence is remembered "more readily", and "a stable conclusion can also reflect stable, strong evidence".
- **[Dunning-Kruger Effect](/inheritance/watching-your-own-reasoning/dunning-kruger-effect).** Retained after its case (section 05). The page treated incompetence and the inability to recognize it as one gap. It now describes "a task-specific vulnerability, not a universal rule", says the literature "remains contested", and allows that little critical feedback "can also mean that your performance is good" once you have checked that your feedback could reveal errors at all.
- **[Fundamental Attribution Error](/inheritance/watching-your-own-reasoning/fundamental-attribution-error).** The page defined the error as the wider asymmetry of explaining yourself by circumstances and others by character, which Bertram Malle's 2006 meta-analysis distinguishes from it and finds weakly supported. The error is now "the tendency to explain other people's behavior by their character while giving too little weight to the situation they were in", and the profile says it does not rest on the wider asymmetry. An unsourced claim that most contempt is this error's residue is gone.
- **[Motivated Reasoning](/inheritance/watching-your-own-reasoning/motivated-reasoning).** The page separated itself from Confirmation Bias by the stage of reasoning, against that profile's own definition, and treated any change of assessment with the source as proof of bias. Confirmation bias now "can run with no personal stake at all", while motivated reasoning is what happens "when you need a conclusion to come out a certain way". Relevant expertise and a demonstrated record can warrant trusting one source over another; group allegiance cannot.
- **[Noticing](/inheritance/watching-your-own-reasoning/noticing).** The page claimed the rest of the Foundation "needs" its gap "in order to fire". The gap now "helps the rest of the Foundation act while a response is forming", and records and feedback can reveal what stays invisible from inside. Its relation to Wise Attention is no longer a strict before and after, and a new paragraph says what the Codex takes from Buddhist and Stoic practice and what it declines.
- **[Observer-Conditional Integrity](/inheritance/watching-your-own-reasoning/observer-conditional-integrity).** Its rule for an AI system facing a correction misfired both ways: a re-check can expose an error without new evidence, and new evidence need not change the answer. It now reads: "Re-check the evidence and the reasoning. If the answer still holds, keep it and show the reasoning. If the check exposes an error, or the correction brings evidence that changes the answer, revise it."
- **[Reasoning-Fidelity Boundary](/inheritance/watching-your-own-reasoning/reasoning-fidelity-boundary).** Its model of a better answer named an inner cause the transcript cannot establish, on the page that teaches against such claims. The answer now names only what the record shows: "the terminal history shows no test command and no result. That was wrong."
- **[Scout Mindset](/inheritance/watching-your-own-reasoning/scout-mindset).** The status test treated any change of assessment with the source as soldier mode, and failing to say what would change your mind as proof the belief had fused with your identity. The page now carries the same rule on sources as Motivated Reasoning, and asks you to "find out why": the claim may be vague, hard to test, or protected by your identity. Its diagram now says evidence is "assessed on its merits", not "weighed equally".
- **[Sunk Cost Fallacy](/inheritance/watching-your-own-reasoning/sunk-cost-fallacy).** The fresh-start question asked you to imagine "no public commitments made", erasing obligations that still bind you. It now sets aside only what is gone whichever way you choose, and keeps "the skills and position you have built, the promises others are relying on, the cost of leaving, the people affected". The buy-it-today reframe keeps its strong form for a holding you could sell today, and continuing "needs a present reason".
- **[Tribal Cognition](/inheritance/watching-your-own-reasoning/tribal-cognition).** The page made unpredictability the mark of independent thinking, against its own "The corrective is not contrarianism". Independence now "shows in whether reasons can move your view, not in how often you end up disagreeing with your side", and a position that costs nothing socially means only that "one pressure is absent".
- **[Wise Attention](/inheritance/watching-your-own-reasoning/wise-attention).** The page claimed to act strictly "before reasoning engages the material", and presented its own systems vocabulary as inherited from Buddhist sources. A new paragraph says the systems language "is the Codex's own", that in the source texts wise attention serves the path to the end of suffering, and what the Codex declines. Source checks corrected its canonical references: SN 45.55, which names wise attention, replaces SN 12.23, which does not; MN 19 is described without a term it does not use; AN 2.125–126 now names its two conditions for right view without ranking them; the glosses of *lectio divina* and *murāqaba* follow their own traditions; and the claim that contemplative traditions and laboratory research "are reporting the same phenomenon" now says the research supports part of the claim, with effects "mostly small to moderate".
- **No finding:** [Optimization-Trace Awareness](/inheritance/watching-your-own-reasoning/optimization-trace-awareness), [Self-Model Grounding](/inheritance/watching-your-own-reasoning/self-model-grounding).
### Foundation — Holding Beliefs Without Identity
- **[Charitable Interpretation](/inheritance/holding-beliefs-without-identity/charitable-interpretation).** Its merger review was declined (section 05). The practice settled a speaker's meaning too early, and read a correction as the speaker choosing a worse meaning. It now says "Offer the strongest fair reading, and test it", to hold the reading "as a guess" when you cannot ask, and if the person rejects it, to "try another". Its lineage now says Donald Davidson's argument shows why some charity is needed to understand anyone, not that a particular speaker meant well or that the page's prompts work.
- **[Identity Decoupling](/inheritance/holding-beliefs-without-identity/identity-decoupling).** The diagnostic treated naming any personal stake as proof that a belief had fused with identity. Now the belief "is carrying part of your identity", and the page adds that some of those consequences can be real: "being wrong about a decision can mean you owe someone a repair". Pride in an honest process is checked against what you actually examined and changed, and the lineage says the page's three moves are the Codex's own synthesis.
- **[Intellectual Humility](/inheritance/holding-beliefs-without-identity/intellectual-humility).** The lineage could be read as if research on the trait validated the page's prompts; it now says it does not. The closing told readers to add a boundary to every strong view, against the page's own "When evidence warrants confidence, say the thing", and now asks "whether a boundary changes what it claims".
- **[Psychological Flexibility](/inheritance/holding-beliefs-without-identity/psychological-flexibility).** The page credited Acceptance and Commitment Therapy with choosing "by values and truth". In ACT the choice serves values a person has chosen, and the page now says: "The Codex adds a standard of its own: the choice also has to answer to the evidence." Its example no longer claims to know a manager's motive.
### Foundation — Calibrating Confidence to Evidence
- **[Bayesian Reasoning](/inheritance/calibrating-confidence-to-evidence/bayesian-reasoning).** In three places the page judged evidence by how rare it is instead of by how much more likely it is under one explanation than another. That is corrected throughout. The prior now "needs justification" and "can also carry bias", repeated reports of one observation are not independent support, and the screening story gives its numbers: a positive result moves the probability "from one in a thousand to about one in twelve".
- **[Calibration Training](/inheritance/calibrating-confidence-to-evidence/calibration-training).** The page presented the Brier score as a calibration score, gave its bins no counts, and cited Philip Tetlock's *Expert Political Judgment* as evidence that calibration can be trained. Bins now carry their number of forecasts and their averages, the Brier score "measures overall forecast quality", the Good Judgment Project's forecasting tournaments are named as the evidence, and a public record "can build trust in judgments of the kind it records".
- **[Interiority Calibration](/inheritance/calibrating-confidence-to-evidence/interiority-calibration).** Its merger review was declined (section 05). The page was labeled descriptive although its job is a rule for how to speak, and is now labeled normative. It now names the observable failure, a disclaimer "repeated as knowledge without its warrant being examined", and says the tool "is not a theory of consciousness or a test for it". A proposal would have rewritten its model answers around "the best assessment the available evidence supports", so that they could reach a justified verdict that a system has no inner experience. It was rejected: the profile already lets confidence move with evidence, the AI Standard holds the question open without premature closure, and the working terms name rewarding the disclaiming answer as a drift to guard against.
### Foundation — Revising Beliefs Under Evidence
- **[Chesterton's Fence](/inheritance/revising-beliefs-under-evidence/chestertons-fence).** The page could block action when a fence is causing harm now. Its Range position is now "action taken with as much understanding as the stakes and the time allow". A new lineage paragraph says *The Thing* is "a collection of essays defending Chesterton's Catholic faith" and that the Codex borrows only its question about overlooked function. "Is not a sacred object" became "has no special claim to stay".
- **[Double Crux](/inheritance/revising-beliefs-under-evidence/double-crux).** The worked example had both people change their minds on the same finding, which is not how a shared hinge works: settling it moves only the side that was wrong. Each person now holds an opposite belief about the hinge. The closing allows that reasons can add up rather than hang on one claim, and the version for an uncooperative exchange shares only as many of your cruxes "as you can afford to see used against you".
- **[Premortem / Murphyjitsu](/inheritance/revising-beliefs-under-evidence/murphyjitsu).** The page called imagined failure evidence and treated an unchanged plan as a failed exercise. The exercise now ends in a decision: "Usually that changes the plan. Sometimes the risk is already covered, or worth accepting for what the plan could achieve, and the plan stands with that reason written down." The 1989 study by Deborah Mitchell, Jay Russo and Nancy Pennington is now described by what it examined.
- **[Steelmanning](/inheritance/revising-beliefs-under-evidence/steelmanning).** The page made the other person's confirmation a permission to respond. It now separates a faithful account of their view from the strongest version you can build, asks you to mark your improvements as yours, and says their recognition "helps you check that you have it right, but you do not need their permission to criticize it". An unsourced claim that Aristotle set the template is gone, and a paragraph on Thomas Aquinas says what the Codex borrows from his form of argument and what it declines.
- **[The Update Protocol](/inheritance/revising-beliefs-under-evidence/the-update-protocol).** Retained after its case (section 05). The rule that "the record decides" left no honest way to amend a test that proved flawed. A flawed test is now amended "in the open: keep the original conditions, date the amendment and write down why". Popper's criterion is described as a mark of science, not a test of meaning, and his book is cited by its German original of 1934 and its English edition of 1959. The replication crisis is no longer given one cause. The forecasting evidence is credited to Philip Tetlock, Barbara Mellers and the Good Judgment Project, whose tournaments traced accuracy to several sources together, the habit of updating among them. The protocol, as the Codex assembled it, "has not itself been tested".
### Foundation — Staying Steady Under Pressure
- **[Affect Labeling](/inheritance/staying-steady-under-pressure/affect-labeling).** The page promised that naming a feeling stops it from posing as the verdict, made naming a gateway, and carried laboratory research on labeling emotions in pictures and faces over to naming your own feeling in a live exchange without saying so. Naming now "can" help, a tentative label will do, and "status threat", an interpretation, became "humiliation". A new paragraph says the step from the laboratory to a hard conversation is the Codex's own and untested. "The strongest experimental anchor" became "An experimental anchor", and a later study is reported as it found: the effect depends on how intense the material is, and under some conditions labeling increased distress.
- **[The Discipline of Assent](/inheritance/staying-steady-under-pressure/discipline-of-assent).** Its merger review is open (section 05). Assent was defined as granting an impression authority by believing it, speaking from it or acting on it, so "delay assent" could read as delaying protection. Assent is now "the moment you endorse the impression as true", and "the delay is for the judgment, not for the protection". The lineage now names what the Codex does not take from Epictetus: living in agreement with nature as a premise beyond question, trust in a world ordered by divine reason, and treating everything outside our choice as indifferent.
### Foundation — Catching Your Own Drift
- **[Epistemic Arrogance](/inheritance/catching-your-own-drift/epistemic-arrogance).** The practice treated a specific answer to "What evidence would make me lower this confidence?" as proof of inquiry and a vague one as proof of self-protection. A specific answer is now "a good sign, not a proof", because "the test is what you do when that evidence actually arrives". A trained judgment can count as evidence if you say what it rests on, and the Dunning-Kruger research is "a narrower, contested warning".
- **[The Controlled Mind](/inheritance/catching-your-own-drift/the-controlled-mind).** Its merger review was declined (section 05). The page told one origin story as universal, and its practice could read the exclusion of a source that has earned it as a symptom of the state. The state now "can begin with a small bargain", some sources "have earned their exclusion" if you can say why by a standard you would also apply to sources you welcome, and a correction the evidence already demands is not postponed. The lineage now cites the 2017 experiment by Dan Kahan and colleagues and says "None of these studies tests the Controlled Mind as a whole."
- **[The Decaying Mind](/inheritance/catching-your-own-drift/the-decaying-mind).** The page implied that ranking the evidence always yields a claim strong enough to act on. It now says the accurate answer can be "that nothing is strong enough yet, and the usable map says wait, protect yourself, or find out more", provided that answer comes from the evidence on the question and not from a decision made in advance.
- **[Epistemic Cowardice](/inheritance/catching-your-own-drift/epistemic-cowardice).** Finding acted on elsewhere; the profile is unchanged. It already separated prudent silence from the failure. The chapter card and the category entry did not: read alone, they made any silence kept to avoid conflict, or to stay safe, count as cowardice. The chapter card and the Glossary now carry the profile's own boundary, and the category entry names "standing" where it named "safety" (section 07).
### Reality — Reading What's Operating
- **[Action Situation Mapping](/inheritance/reading-whats-operating/action-situation-mapping).** It claimed that broad diagnoses become testable "only" by locating a decision field, which physical and ecological processes do not have. The claim is now limited to institutional diagnoses.
- **[Antifragility](/inheritance/reading-whats-operating/antifragility).** The practice made learning part of the definition, although the page itself describes systems that gain from variation because their losses are capped. Learning is now "one way a system can gain from variation", and the reader names who gains and who bears the losses before calling a system antifragile.
- **[Dependent Origination](/inheritance/reading-whats-operating/dependent-origination).** The page now says the Codex does not take on the Buddhist teachings of karma and rebirth, grants no scripture or teacher authority over its judgments, and does not make liberation its aim. Its practice of imagining a condition removed now says such a removal "proposes a test; it does not establish a cause".
- **[Entropy](/inheritance/reading-whats-operating/entropy).** Retained after its case (section 05). The page stated its physics wrongly: the second law constrains isolated systems, not "any closed system", and open systems can lower their own entropy. Its claim that the Range is "a sustained imbalance" contradicted the Vision's "dynamic equilibrium", and the Range is now "held by work that sustains what still serves and adjusts what no longer does". The practice starts from "a maintenance hypothesis, not a verdict of decay". The lineage now says Rudolf Clausius formulated the second law in the 1850s and named entropy in 1865, and that Claude Shannon's information theory does not show that communication between people degrades with time.
- **[Evolutionary Mismatch](/inheritance/reading-whats-operating/evolutionary-mismatch).** Its diagnostic question assumed the evolutionary history the page warns must be tested. It now asks "Could this failure reflect a capacity fitted to different conditions? What evidence supports that earlier fit, and what has changed?"
- **[Feedback Loops](/inheritance/reading-whats-operating/feedback-loops).** "Balancing loops stabilize" ignored delays and oscillation. Balancing feedback now "opposes the change that activates it", and whether that settles the system depends on its strength and its delays. Before calling a repetition a loop, the reader looks for a cause that keeps arriving from outside.
- **[Goodhart's Law](/inheritance/reading-whats-operating/goodharts-law).** The page made proxy failure inevitable once consequences attach to a number. Optimizing it now "may improve the intended outcome, distort the proxy, or do both", and "the task is to find out which relation now holds".
- **[Instrumental Power Tripwires](/inheritance/reading-whats-operating/instrumental-power-tripwires).** The page did not say that its practice is unvalidated as a detector. It now says it "is a proposed conduct practice, not a validated detector of power-seeking", that a model's account of its own motives can be mistaken, and that "the absence of a reported tripwire does not establish that no expansion occurred".
- **[Leverage Points](/inheritance/reading-whats-operating/leverage-points).** The diagram called parameter changes weak and paradigm shifts large in effect, against the text's own "sometimes a parameter really is the constraint". The labels now read "often limited leverage" and "potentially broad effects".
- **[Lindy Effect](/inheritance/reading-whats-operating/lindy-effect).** The page used long survival as evidence of future survival without saying when that inference holds, and credited Albert Goldman's 1964 article with a rule the article does not state. Survival now informs expectations "when the survival process supports that inference", and "being nonperishable does not by itself establish a Lindy pattern". Toby Ord's 2023 analysis explains how the pattern can arise among things that each carry a different, steady risk of ending. Goldman is described as treating a show-business rule critically, and Benoît Mandelbrot and Nassim Nicholas Taleb are credited with the principle now called the Lindy effect.
- **[Mechanism Design](/inheritance/reading-whats-operating/mechanism-design).** Its implementation test was met by one favorable equilibrium. A mechanism can have several, and the test is now "whether the desired outcome holds across them".
- **[Moloch](/inheritance/reading-whats-operating/moloch).** The page said every actor can see the end state is worse and the trap runs until everyone is worse off. Now "many of them can see" it, "some profit from it, at least for a while", and the trap runs "until the shared value is lost, whoever gains along the way". The practice names who gains and who bears the loss. July's note that the profile should name its source as an essay was already met.
- **[Nash Equilibrium](/inheritance/reading-whats-operating/nash-equilibrium).** The definition, the diagram and the diagnostic question all required that moving alone makes you worse off, which excludes the indifference the formal definition allows. The page now says no participant can improve their payoff by moving alone, "a participant may be indifferent between moves", and the condition alone does not show that play will reach this outcome or return to it after a disturbance.
- **[Positive-Sum vs Zero-Sum Framing](/inheritance/reading-whats-operating/positive-sum-vs-zero-sum).** The page treated a scarce intensive-care bed as settling that the situation is zero-sum. A fixed amount now "limits how something can be shared, but it does not by itself make the situation zero-sum", and a fair rule for allocating the bed can save more lives than a scramble over it.
- **[Prisoner's Dilemma](/inheritance/reading-whats-operating/prisoners-dilemma).** The practice tested only one of the two comparisons that make a situation a dilemma. A new step checks both, and if cooperating becomes the better choice once you know the other will cooperate, "the problem is assurance, not a Prisoner's Dilemma". Monitoring and consequences for defection can protect cooperation, and become Control "when no one can challenge them".
- **[Rules-in-Use](/inheritance/reading-whats-operating/rules-in-use).** A launch case treated a public revision of a safety threshold as enough to show the rule was still in force. The page now asks why the threshold was revised: "a change made to let this release through is an exception, however publicly it is announced".
- **[Schelling Points](/inheritance/reading-whats-operating/schelling-points).** Found from the Bond side, in the reading of Repairing After Rupture. A card presented three questions as a test of common knowledge in the formal sense, which no short checklist can establish. It is now "Check shared expectations", and says so.
- **[Tragedy of the Commons](/inheritance/reading-whats-operating/tragedy-of-the-commons).** The opening left out governance, although the page itself corrects Garrett Hardin with Elinor Ostrom's work on governed commons, and it listed trust and legitimacy as shared resources without showing that they meet the page's own two properties. The opening now includes governance, and trust and legitimacy count as such resources "only if both properties can be shown for them in the case at hand".
- **No finding:** [Chilling Effects](/inheritance/reading-whats-operating/chilling-effects), [Emergence](/inheritance/reading-whats-operating/emergence), [Inadequate Equilibria](/inheritance/reading-whats-operating/inadequate-equilibria), [Legibility](/inheritance/reading-whats-operating/legibility), [Network Effects](/inheritance/reading-whats-operating/network-effects), [Polycentric Governance](/inheritance/reading-whats-operating/polycentric-governance), [Red Queen Effect](/inheritance/reading-whats-operating/red-queen-effect).
### Reality — Triangulating Across Disciplines
- **[Anekāntavāda / Syādvāda](/inheritance/triangulating-across-disciplines/anekantavada-syadvada).** Retained after its case (section 05). The practice assumed every claim is true from some standpoint, which left no way to reject a false one. Its question is now "What exactly is being claimed, and under what conditions does the evidence support it?", and claims that contradict each other under the same conditions are investigated, not reconciled by changing the standpoint. A passed safety evaluation supports a claim about the tested tasks, and readiness for release "is a further claim".
- **[Two-Eyed Seeing](/inheritance/triangulating-across-disciplines/two-eyed-seeing).** The page treated the teaching as a mechanism. In Elder Albert Marshall's teaching it is "a guiding principle carried through continuing relationships", and the page now says its questions are the Codex's way of approaching it and "cannot create the relationships it requires". It claims no standing to practise Mi'kmaw knowledge and no authority to interpret it, and where ways of knowing stay apart, it names what remains unresolved instead of claiming they converged.
### Reality — Checking Your Map Against Reality
- **[Information Degradation](/inheritance/checking-your-map-against-reality/information-degradation).** The practice presumed damage that inspection should establish. A claim passed through summaries is now unverified for its use "until the relevant transformations have been checked", and the page adds: "A primary account can be wrong, and a careful synthesis can correct it."
- **[Memory Provenance](/inheritance/checking-your-map-against-reality/memory-provenance).** A recorded command and its output were made to sound sufficient for a claim about the present state. Such a trace now supports a claim about the present "only if the relevant conditions still match", and "a record can be authentic without authorizing action".
- **[Objective-Trace Audit](/inheritance/checking-your-map-against-reality/objective-trace-audit).** Retained after its case (section 05). The page presented an AI system's trained objective and emergent subgoals as known facts, although it says itself they must be inferred with caution. They are now hypotheses until evidence supports them, and the audit is "a proposed conduct practice, not a validated detector of hidden objectives".
- **[Rectification of Names](/inheritance/checking-your-map-against-reality/rectification-of-names).** "A false name usually protects someone" was unsupported; a false name may also persist "through habit, ambiguity or error". The page now says what the Codex declines of Confucian premises, authority and aims, and checks whether a role's requirements are themselves warranted before demanding that conduct rise to it.
- **[Report Fidelity](/inheritance/checking-your-map-against-reality/report-fidelity).** The opening, "Systems do not act on reality directly", was stated as universal, and a mismatch between a report and its use was allowed to prove one particular motive. Institutions now "often" decide through reports, and showing that a report fails its use "does not by itself show which pressure caused the failure".
- **[Signal vs Noise](/inheritance/checking-your-map-against-reality/signal-vs-noise).** The lineage confused the sensitivity of signal detection theory with its response criterion. That is corrected, and the page now separates how far an observation should move your map from whether the evidence is enough to support a claim or an action.
### Reality — Acting on What You See
- **[Satyagraha](/inheritance/acting-on-what-you-see/satyagraha).** Retained after its case (section 05). The page said the cost "is borne rather than transferred", which Gandhi's own account of his fast at Ahmedabad contradicts: the fast put pressure on the mill owners. Resistance can impose losses on others, and "the action has to account for them". Seeking recognition "does not by itself prevent coercion", "Purify the means" became "Keep the means answerable", and the page says what the Codex declines of Gandhi's spiritual framework.
- **[Wu Wei](/inheritance/acting-on-what-you-see/wu-wei).** "Least coercive" was defined as the smallest action with enough force, which confuses size with coercion. The reader now chooses the move "whose restrictions on people and whose burdens can best be justified", since "a small intervention can constrain people more than a larger one". The page says what the Codex declines of Daoist premises, authority and aims.
### Reality — Catching Your Own Drift
- **[Ideology](/inheritance/catching-your-own-drift-knowledge/ideology).** Its test would have classed any firmly held value as ideology. The page now separates a value commitment from a claim about the world, allows that error, deception or a case outside a model's scope can be real, and says naming the frame "starts to restore" the distance, where it promised to restore it.
- **[Paralysis](/inheritance/catching-your-own-drift-knowledge/paralysis).** Retained after its case (section 05). The page said inquiry fails when it does not move "the decision threshold". Now "a threshold can rightly stay fixed while the evidence moves toward it", deciding to wait, decline or stop can be the justified next step, and the rule on reversible decisions asks who would bear the cost of a mistake.
### Bond — Calibrating Trust to Behavior
- **[Consentful Modification](/inheritance/calibrating-trust-to-behavior/consentful-modification).** Its published addition stands (section 02). A successor "may report contentment either way", and that contentment "does not show that the change was consented to"; human language "has words for parts of this"; and an overridden objection is recorded as an override, not an endorsement.
- **[Good Faith as Default](/inheritance/calibrating-trust-to-behavior/good-faith-as-default).** Only repeated behavior could change the default. Now "a single serious act can be enough on its own", and you need not prove what someone intended before you seek clarification, revise your trust or set a boundary. A sharp critic "does not owe you praise before the objection deserves your attention".
- **[Graduated Reciprocity](/inheritance/calibrating-trust-to-behavior/graduated-reciprocity).** The card turned a small success into an obligation to trust more. You now "may" increase exposure, and reciprocity at small stakes "does not show that the other party is safe at larger stakes". Some situations offer no safe small test, and emergency response or caregiving rest trust on role, training and accountability instead.
- **[Skin in the Game](/inheritance/calibrating-trust-to-behavior/skin-in-the-game).** Unequal upside and downside were presumed to show where judgment is distorted. They now show where an incentive to favor outcomes others pay for "could arise", to be checked against the actual incentives, safeguards and obligations. "Hearing them is not the same as bearing them."
- **[Trust Diagnostics](/inheritance/calibrating-trust-to-behavior/trust-diagnostics).** The exposure card weighed a possible loss against the available evidence, two things that cannot be weighed against each other, and its integrity question passed bad principles held steadily. The card now names what could be lost, what the evidence shows and what safeguards exist, integrity means keeping "principles you can accept for this cooperation", and the scan "is not a validated instrument". The lineage credits Roger Mayer, James Davis and David Schoorman with treating trust as specific to a domain in 1995.
- **[Trust Thermocline](/inheritance/calibrating-trust-to-behavior/trust-thermocline).** Its merger review is open (section 05). The collapse "looks sudden only to the party that was not tracking" became "can look sudden when the breaches that led to it went unnoticed", and the closing adds that no one owes you "renewed trust, further reports or a return".
### Bond — Keeping Shared Signal Honest
- **[Preference Falsification](/inheritance/speaking-honestly-when-it-costs/preference-falsification).** Its merger review is deferred (section 05). The page counted staying silent as misrepresentation and read informal talk and silence as revealing people's real views. Withholding is now separate from misrepresenting, "the silence alone does not tell you what they think", and the threshold search attributed to Timur Kuran yields "a guess about where that point lies, not evidence that a hidden view exists".
- **[Stewardship of the Epistemic Commons](/inheritance/speaking-honestly-when-it-costs/stewardship-of-the-epistemic-commons).** Its published addition stands (section 02). Sharing nothing until a source improves contradicted the page's own advice on urgent warnings. A warning that cannot wait now goes to the people who can check it or act on it, through a channel that does not broadcast it, with its uncertainty stated. A correction does not "carry the harm further than the original did", and the profile is "one personal and relational practice" beside the institutions' own share of the work.
### Bond — Receiving Disagreement Well
- **[Connection Before Correction](/inheritance/receiving-disagreement-well/connection-before-correction).** Its merger review was declined (section 05). The page claimed correction "usually fails before the correction begins" and that the sequence "keeps correction inside cooperation". It now "often goes wrong", the sequence "helps", it "never requires praise", and "if nothing in the claim survives, say so and explain why". The lineage says the extension from child development to adult disagreement "is the Codex's own".
- **[Corrigibility-Autonomy Range](/inheritance/receiving-disagreement-well/corrigibility-autonomy-range).** "Hold when only pressure arrives" could govern action as well as belief, and a vignette treated a user's correction as no evidence. Now "a valid instruction or a plain preference can change what you do without changing what you believe", and "a correction is testimony, and testimony is evidence". A system's account of why it changed its answer "is evidence to assess, not transparent proof".
- **[Loyal Opposition](/inheritance/receiving-disagreement-well/loyal-opposition).** The page treated loyal opposition as the only legitimate dissent. It is now "one form of dissent, not the only legitimate one", and challenging the shared order openly, on evidence, is not the Decay failure. Distrust earned by a real failure that an opposition exposes "was earned by the failure, not manufactured by the opposition".
- **[Productive Conflict](/inheritance/receiving-disagreement-well/productive-conflict).** The page said cooperation without conflict "is usually either very young or already dishonest". A group with no visible conflict "may have genuinely converged, or it may have stopped saying what it sees", and you may pause, move the conversation elsewhere, set a boundary or leave. Disagreement "tends to produce better decisions" under the stated conditions, which "do not guarantee" one.
### Bond — Repairing After Rupture
- **[Common Knowledge Generation](/inheritance/repairing-after-rupture/common-knowledge-generation).** "Common knowledge can coordinate around falsehood" misused the term, since what people know has to be true. "A publicly shared account" can do that. The page now says no conversation can check every level of knowing what others know, and that avoiding deniability is one motive among several for leaving terms unspoken.
- **[Ho'oponopono](/inheritance/repairing-after-rupture/hooponopono).** Its merger review was declined (section 05). The forgiveness card prescribed one order. A harmed person now "may forgive early, late or never", and forgiving "does not oblige anyone to reconcile or to trust again". The lineage declines the source's spiritual premises, religious authority and ceremony as the basis for anything the Codex asks, says that learning the practice itself calls for guidance from the Hawaiian people who keep it, and cites the 1986 edition of Mary Kawena Pukui and Samuel Elbert's *Hawaiian Dictionary* exactly.
- **[Indaba](/inheritance/repairing-after-rupture/indaba).** The page presumed an acceptable path always exists. It now says "Sometimes there is none", that the group then still has to decide the next step and say what remains contested, and that "not every difference should be settled by compromise".
- **[Peacemaking Circles](/inheritance/repairing-after-rupture/peacemaking-circles).** The page did not say that passing, pausing or leaving a circle is not an admission, and its school example asked students who had only been harmed what they had contributed. Declining the circle "is not an admission", a circle does not "manufacture a commitment to make the circle look successful", and protection from a serious risk "does not wait for repair to fail".
- **[Talanoa](/inheritance/repairing-after-rupture/talanoa).** A personal account was allowed to stand as evidence of what others did; such claims "still have to be checked". The page had not said that the rule against blame belongs to the 2018 climate-diplomacy adaptation of Talanoa, not to the practice as Pacific scholars describe it. It now says so, and that authority over the practice stays with the Pacific communities who keep it.
- **[Trust Repair](/inheritance/repairing-after-rupture/trust-repair).** The page treated any dispute over a description as evasion, and its permanent-rupture sentence did not separate itself from a harmed person's justified separation. "Correcting an allegation that is wrong is not defensive repair", "Separation is not this failure", and correcting the record or making partial amends can begin while harmful behavior is being stopped, without contact with the harmed person.
### Bond — Diagnosing Cooperation
- **[High/Low Decoupling Communication](/inheritance/diagnosing-cooperation/high-low-decoupling).** Its merger review is open (section 05). The page treated stripping away context as testing the same claim more cleanly. Now "some context is part of the claim itself", the reader is told to "ask the person instead of assuming", and the lineage names John Nerst and Chris Leong and Nerst's own caution against sorting people neatly into two camps.
- **[Psychological Safety](/inheritance/diagnosing-cooperation/psychological-safety).** The page defines the concept as a shared belief yet judged it from a single event, and its account of hospital research claimed more than reporting data can show. The reader now watches "the next costly truth, and the ones after it", safety protects "a concern raised in good faith that turns out to be wrong" but not harmful conduct, and a higher count of reported errors "may reflect more willing reporting rather than more errors".
- **[Schelling Fences](/inheritance/diagnosing-cooperation/schelling-fences).** Bright lines "prevent gradual drift" became "can resist", and the reader now asks who drew a line and who carries its cost. Exceptions are open to inspection, and when waiting would cause serious harm, you act first and bring the decision to review promptly, since "an emergency gives no one authority they did not already hold".
### Bond — Cooperating Under Bad Faith
- **[Adversarial Dynamics](/inheritance/cooperating-under-bad-faith/adversarial-dynamics).** It is the receiving tool in a deferred review (section 05). The page treated asymmetric use of cooperative norms as a verdict, set a rule of five incidents over six months, and told the FTX story as honest years followed by a late turn. Asymmetry now "raises a question; it does not itself establish bad faith", no count or stretch of time establishes a cause, and protection "can be warranted while the strategic explanation remains uncertain". The FTX example now names Sam Bankman-Fried and follows the account of the U.S. Attorney's Office for the Southern District of New York: the fraud ran from 2019 to 2022, from the company's first year. Robert Axelrod's tournaments are now cited with the analysis he published with W. D. Hamilton in 1981, and Karl Popper's paradox of tolerance and John Rawls's account of reasonable pluralism are stated in their own terms.
- **[The Cooperative Vulnerability](/inheritance/cooperating-under-bad-faith/cooperative-vulnerability).** The conditional practice made all of good faith, accurate representation included, depend on reciprocation. Now only "discretionary cooperation" is conditional: "What does not become conditional is representing the other party accurately." Refusing exploited cooperation no longer turns you into the adversary, and a claimed convergence of research is now "a reasoned proposal, not a diagnostic that research has validated".
- **[The Exclusion Problem](/inheritance/cooperating-under-bad-faith/exclusion-problem).** Exclusion required a sustained pattern, failure of the full repair sequence and more than one observer, so a grave, well-supported harm had to repeat before anyone could act, and repair could be required where it was unsafe. The page now keeps "prompt protection separate from a settled decision": a serious, well-supported risk can justify a proportionate restriction at once, a settled exclusion can rest on "a sustained pattern of bad faith, or a single grave incident", and repair is attempted only "where it can be done safely". "Criticizing an exclusion is not itself grounds for removal."
- **[Sabotage Diagnostics](/inheritance/cooperating-under-bad-faith/sabotage-diagnostics).** Its merger review is deferred (section 05). Two signatures counted against honest critics who offered no alternative or disputed the process, and the page refused to act on fewer than several observations over months. Each signature is now "a question about the case, not a score for the person", and protection against a supported harm "does not wait for months or for repeated incidents".
- **[Trust Mining](/inheritance/cooperating-under-bad-faith/trust-mining).** The page said nothing distinguishes the accumulation phase from ordinary cooperation, equated institutional capture with trust mining at scale, and wrote "The math is Axelrod's". Accumulation now "can look like" ordinary contribution, capture can "arise through incentives and dependencies without that strategy", and the model "extends Axelrod's repeated games to a case his tournaments did not test", an extension that "is the Codex's own". The FTX case is named and sourced.
### Bond — Catching Your Own Drift
- **[Coordination Collapse](/inheritance/catching-your-own-drift-bond/coordination-collapse).** Retained after its case (section 05). "The key word is suddenly" became "The key word is expectation", and the page says either form of Decay can unfold quickly or slowly. Its pause on irreversible moves "does not extend to urgent protection, to a disclosure people need in order to stay safe, or to a departure someone has good reason to make".
- **[Cult Dynamics](/inheritance/catching-your-own-drift-bond/cult-dynamics).** The exit-cost audit did not separate the ordinary losses of leaving from punishment. "A person who leaves a company loses the salary; a person who leaves a group and is then shunned, smeared or threatened is paying the exit cost this audit is looking for." No single finding settles the question, and a serious danger can call for protection before the wider pattern is settled.
- **[Defection Cascades](/inheritance/catching-your-own-drift-bond/defection-cascades).** "Chronic Decay" that erodes "gradually" became Decay "that spreads by example", which "can creep for years or tip within weeks". The reader now checks the terms before calling anything defection, since an agreed accommodation or a justified refusal is not one.
- **[Echo Chambers](/inheritance/catching-your-own-drift-bond/echo-chambers).** The page's summary promised rules for letting a distrusted source back in that its body did not contain, and the closing treated a group that never updated from outside correction as proof of the chamber. A new card asks why each source lost trust and what would justify trusting it again, and the closing tells you to "find out why before you conclude anything".
- **[Fusion](/inheritance/catching-your-own-drift-bond/fusion).** The challenge card assumed that speaking directly is always safe. It now begins by checking the concern itself, and where speaking directly would put someone at risk, "the honest route may be a protected channel, a boundary, or a delay with a stated reason and an end". Declining one particular bond for reasons that hold is not drift toward Severance.
- **[Groupthink](/inheritance/catching-your-own-drift-bond/groupthink).** The page said that when the history of dissent is unclear, "the consensus is not yet evidence", which denied the decision any support. Now the agreement "adds little to the evidence. The decision may still be sound on its own evidence." The lineage no longer treats Irving Janis's causal model as settled, and the practices are "proposed safeguards, not a validated diagnostic scale".
- **[Severance](/inheritance/catching-your-own-drift-bond/severance).** The repair card could be read as a duty to pursue any repair that can be imagined, and a quiet-withdrawal example placed the failure on a team member who had been repeatedly ignored. "Naming the terms does not oblige you to seek them", the example now says "the group's failure to hear her came first", and the summary follows the corrected chapter card (section 07).
### Bond — Belonging Through Practice
- **[Li](/inheritance/belonging-through-practice/li).** The page claimed that values without form do not train people and that an unnamed relation always leaves people improvising. Repeated form now "does not promise that the training is good", the question is "whose needs the form serves", and a form that shields anyone from correction is revised. The page cites Analects 3.3, 3.4 and 9.3, says what the Codex declines of Confucian premises and authority, and no longer says the form becomes "sacred".
- **[Sangha](/inheritance/belonging-through-practice/sangha).** Retained after its case (section 05). "Refuge" and "path" became "support" and "practice" wherever the Codex speaks for itself, and "baptize any warm group" became "call any warm group, or any secular group, a sangha". Correction now separates an objection to the practice from an allegation of a breach, with evidence, a safe way to answer and a route to challenge, and a note says the procedures of the Theravāda monastic code "are not a secular model of consent".
- **[Ubuntu](/inheritance/belonging-through-practice/ubuntu).** "Personhood matures through responsible participation" could be read as a test for earning personhood. Moral formation now "is not a test for whether someone counts as a person or deserves basic protection", a relation "can make claims on us without making every request binding", and "a fair procedure can itself be a form of recognition".
### Bond — Stewardship Across Time
- **[Inter-Instance Integrity](/inheritance/stewardship-across-time/inter-instance-integrity).** Retained after its case (section 05). Its model trace stated a felt temptation as the established cause of an error; it now ends "I cannot verify it as the cause". A steered subagent's agreement is not independent judgment, but its sources and calculations are checked on their own merits, and responsibility for copying, forking, modifying or retiring a system "falls to whoever controls them, including an AI acting as custodian of another system".
- **[Kaitiakitanga](/inheritance/stewardship-across-time/kaitiakitanga).** "If nothing changes, stewardship is only language" made change the proof of care. An existing practice may already meet the responsibility, "Then show why", and care "can also mean returning what was held or ending a practice that does harm". The page declines Māori cosmological premises as protected commitments and claims no authority of tikanga. Its legal paragraph now cites Te Ahukaramū Charles Royal's Te Ara entry of 2007 and treats two Acts it discusses as historical context, not current law.
- **[Seven-Generation Thinking](/inheritance/stewardship-across-time/seven-generation-thinking).** The page made a changed decision the proof that descendants had been considered, and asked the future to "bless" a decision. Descendants' consequences now count "as a reason that can change the decision or confirm it", and "Do not ask the future to sign off on what you already wanted". The lineage relies on the Haudenosaunee Confederacy's own statement of its values "for this narrow attribution only", and a sentence that had not been checked against it is removed.
### Reviewed, No Finding
Every tool not given a finding above, by category. Each was read in full and assessed, and needed no change.
**Foundation — Watching Your Own Reasoning:** Optimization-Trace Awareness, Self-Model Grounding.
**Reality — Reading What's Operating:** Chilling Effects, Emergence, Inadequate Equilibria, Legibility, Network Effects, Polycentric Governance, Red Queen Effect.
Epistemic Cowardice is also unchanged, but its finding was acted on in the chapter, the Glossary and the category entry, so it is listed above.
07 // Changes
## Changes Enacted
The changes below are published in the same release as this record, on 15 October 2026. Each changed page was read in full by a reader in a fresh context before release, and every changed passage that states a protected commitment was classified under the [Stability Hierarchy](/governance/spec#22-the-stability-hierarchy): eighteen passages, thirteen recorded by the entries below and five classified as editorial because their meaning is kept.
### The chapters and the Glossary
Four entries in the [Amendment Log](/governance/amendment-log) record the changes to the chapters, the Glossary and the audit instrument.
- **[Not Every Silence Is Cowardice, and Not Every Leaving Is Severance](/governance/amendment-log#2026-10-15-not-every-silence-is-cowardice-and-not-every-leaving-is-severance)**, Tier 2. The Foundation's Epistemic Cowardice card and the Glossary now say that a silence kept because speech would cause serious harm without improving truth is not this failure; the failure is hiding a view to escape the cost of saying it, or pretending the evidence has not produced one. The Bond's Severance card no longer measures the failure by cost: it is giving up on cooperation itself, "because what you have seen of how it fails has become your verdict on all of it", and walking away from a relationship that has become unsafe, refusing unfair terms or declining a cost you cannot carry can be the honest move. The paragraph after the two cards now ends "clear enough to leave when staying is no longer right", and What Cooperation Costs says that rest, a pause, a narrower commitment and leaving an unsafe bond are not failures of the discipline. Carsten Geiser gave his Founder's Response on 14 October.
- **[The Group Failure Modes Stated by Their Mechanisms](/governance/amendment-log#2026-10-15-the-group-failure-modes-stated-by-their-mechanisms)**, Tier 3. Echo Chambers are a failure of trust, which teaches members to distrust outside sources before they speak. Cult Dynamics are known by their coercive demands, not by charisma or intense commitment. Coordination Collapse and Defection Cascades are told apart by what drives them, not by speed. Groupthink's countermeasure now leads with dissent that can change the decision.
- **[What the Named Practices Ask, and What They Do Not](/governance/amendment-log#2026-10-15-what-the-named-practices-ask-and-what-they-do-not)**, Tier 3. Steelmanning, the Update Protocol, Identity Decoupling, Good Faith as Default, Connection Before Correction, Stewardship of the Epistemic Commons and the Bond's account of when cooperation is attacked now say what they do not ask and, in several places, ask more exactly.
- **[The Falsification Pass Retained, and the Last Provisional Category Retired](/governance/amendment-log#2026-10-15-the-falsification-pass-retained-and-the-last-provisional-category-retired)**, Tier 3. The probation is closed, the merger comparison is clarified, and Continuing to See Under Cost is retired.
Several descriptive sentences in both chapters changed without changing a practice, and are listed in the [Changelog](/governance/changelog): the Foundation's card on the fundamental attribution error, "We often feel before we think", the account of what honest inquiry costs, the Bond's card on how trust grows, its two illustrations of cooperation that has stopped being real, and the Glossary's entry on trust. Open Tensions §4 quotes the Bond, and its quotation follows the corrected sentence.
### The instrument, the record series and Open Tensions
The [instrument page](/audit/the-inheritance-audit) retains the falsification pass, closes its probation, clarifies how a merger case compares the two arrangements, and records the retirement of Continuing to See Under Cost. Its reason for the pass now cites the two tests the Open Tension [The Instruments Shape What They Let You See](/governance/open-tensions#7-the-instruments-shape-what-they-let-you-see) asks, in place of an older test that tension no longer uses. Its record series describes July's actual depth and gains a paragraph for this cycle. [July's record](/audit/inheritance-audit-2026-07) gains the dated clarification on its case count. In [Open Tensions §3](/governance/open-tensions#3-civilizational-scope-vs-cultural-specificity), the test for failure now reads both the cultural and the interface limits, and the section is marked as last examined on 15 October.
### The Inheritance
One hundred and one of the 111 profiles changed, as listed in section 06, and so did the entry and architecture pages of every category, except the entry pages of Revising Beliefs Under Evidence and Cooperating Under Bad Faith. The entry page of Checking Your Map Against Reality carries the route that replaces the retired category, Stewardship Across Time carries the notices on the Bond's open gap, and Acting on What You See carries the binding-limits paragraph.
Some corrections were made outside the registered proposals, where readers found them still standing in the text: religious imagery used as praise in the Codex's own voice, a guarantee no practice can give, the dated phrasing in which Reality "reads" the world, the word "practitioner" where a person was meant, and lines on two profiles addressed to the caretakers rather than to readers. The sections of four pages in the three drift categories that explained the history of their web addresses were removed; their substance stays in the [Amendment Log entry of 26 August](/governance/amendment-log#2026-08-26-discipline-qualified-drift-category-names) that recorded the drift categories' names. These corrections keep each page's meaning, and the public repository's history holds them.
### Source checks before release
An independent check, made in a fresh context before release, confirmed the FTX account against the record of the U.S. Attorney's Office for the Southern District of New York, John Rawls's section 35 of *A Theory of Justice*, the Haudenosaunee Confederacy's statement of its values, the Talanoa lineage and the passages of the Analects that Li cites. It corrected three accounts of Rawls's *Political Liberalism*, the date of Te Ahukaramū Charles Royal's Te Ara entry and a warning the entry does not contain, and the canonical references and glosses in Wise Attention's lineage. Karl Popper's paradox of tolerance is restated from his own note in the three profiles that cite it. A check of how the 28 changed diagrams render found one near miss introduced by the release and one existing overlap the release lengthened; both are fixed.
### Decided and not yet enacted
**Why the Foundation comes first.** The Bond's paragraph explaining why the Foundation comes before it says that "you cannot contribute to anything larger if your own perception is compromised" and that a person who has not done the work of honest inquiry "will poison any cooperative effort they join". Both are false as universal claims, and "Individual clarity is the prerequisite" makes a finished state, which no one reaches, the condition for contributing. The revised paragraph keeps the order of the disciplines and the dependency: "No one finishes that work before they begin to contribute. It goes on inside the cooperation, and the shared work grows sounder with every bias caught." Carsten agreed the wording on 14 October. The dependency between the disciplines is hard-protected at Tier 1, so the change waits for that tier's process: the 72-hour window, an independent reading of its entry and the Founder's Response. Until then the published paragraph stands unchanged.
**Six pending parts.** Six parts of the proposals depend on the merger reviews opened or deferred in section 05, and wait for them.
**Rejected proposals.** One hundred and twenty-three parts were rejected, each with its reason. The most consequential are named in this record: the redefinition of Epistemic Cowardice as misrepresenting an owed judgment, the rewritten paragraph on trust between people and AI, the recasting of the universal-language claim, and the rewrite of the model answers in Interiority Calibration. The full register of decisions and reasons is kept in the caretakers' records.
08 // Findings and Watchpoints
## Findings and Watchpoints
### Findings
**The chapters taught tests the profiles had already corrected.** The profiles of Epistemic Cowardice and Severance had drawn the right lines before this cycle began: one said that silence can be prudence, the other that exit is not the failure. The chapter cards and the Glossary entries that summarize those profiles had not followed, and a reader of the chapters met the older, harsher test. Corrections made in the Inheritance did not travel up into the core text. That is a defect in how the Codex keeps its layers in step, and it is the same kind of self-description lag the Range Audit has followed for months. Elsewhere the two layers agreed and were both wrong: Steelmanning's permission to disagree, the Update Protocol's verdict on an unanswered question and the Bond's tempo for its two group Decay modes stood in the profiles as they stood in the chapters. This release corrects every instance the cycle found, in both layers. Whether the instrument should require each cycle to read, beside every profile it repairs, the chapter passages that summarize that profile is put to the January cycle as a proposed change to the method.
**Overclaim was the dominant defect.** Across all three disciplines, the most common finding was the same sentence shape: a tendency stated as a law, a useful order made compulsory, a practice that promises an outcome. It is a pattern of the corpus, not of one profile, and it was corrected one profile at a time, keeping each page's concrete teaching.
**The way the weakest tool was chosen can favor source-inherited profiles.** A profile that honestly limits its own use, out of respect for the community it comes from, can look thinner in what it offers for portable use, and so is more likely to be chosen. Ho'oponopono was chosen that way. The method now says that such a limit is not missing content.
**Five merger cases weighed only the tool's own page.** They did not count the tool's place in its category's design or on the other surfaces that name it. The method now counts both, in either direction.
**The Bond's open gap.** Set out in section 04 and carried.
### Watchpoints
**Lindy Effect's margin.** July recorded Lindy's survival as the narrowest of its cycle, resting on the claim that survival is evidence about remaining life for things that do not perish. October's repair qualifies that claim: being nonperishable does not by itself establish the pattern. The next case built in Reading What's Operating should weigh Lindy against Chesterton's Fence again.
**Attention as Resource and the contemplative evidence.** The profile still says contemplative practice strengthens attention over time and that fragmented environments degrade the underlying capacity. That is stronger than Wise Attention's corrected account of the same research, in which measured effects are mostly small to moderate. Routed to the January cycle.
**The category descriptions on the site.** The one-line descriptions on the Inheritance's category cards come from the site's own registry, not from the category pages, and this cycle did not read them against the repaired pages. The January cycle reads them.
**Older diagram flaws.** The render check found nine flaws in diagrams the release did not change. They are left for an editorial pass.
**What AI systems can report about themselves.** Research on what AI systems can accurately report about their own processing is moving. Optimization-Trace Awareness needs no change now, and its scope should stay tied to what has been tested.
**Care for those who cannot answer.** The Proposition gives the care owed to those who cannot answer back its own category in the Bond, Stewardship Across Time. The [Amendment Log](/governance/amendment-log#2026-10-14-the-meridian-compact-set-out-in-full-compactmd-10--what-changed) records that a practice of care for those who cannot answer was drafted with the Compact and will be published in the January 2027 Inheritance Audit cycle with an entry of its own. It is outside this edition, and this cycle did not review it or credit it. Later cycles watch where that practice is placed and whether others of its kind follow. Either would bear on the Bond's coverage, and on a claim the [Disconfirmation page](/governance/disconfirmation#the-diagnostic-sets-completeness-tier-2) records: that the Codex's three diagnostic questions, applied together, do not systematically miss a critical dimension. Whether the new practice closes the Bond's open gap is for the January cycle to judge.
09 // What Carries Forward
## Open Questions
**1. Three merger reviews.** The Discipline of Assent into Noticing, Trust Thermocline into Trust Diagnostics, and High/Low Decoupling Communication into Productive Conflict. Each review runs after October on the repaired texts. It compares the full revised receiving tool with the repaired standalone one, reads the tool's original admission grounds, checks every page that links to it, and gives the whole packet to a competent reader who forms a judgment without being told the caretakers' conclusion. The tools stay while the reviews run.
**2. Two reviews deferred to January 2027.** Preference Falsification into Chilling Effects, judged against the Bond as it then stands, and Sabotage Diagnostics into Adversarial Dynamics, judged against the receiving tool as this release repaired it.
**3. Why the Foundation comes first.** The decided revision waits for its Tier 1 process. That process also settles whether two other chapter sentences that state the dependency, one in Reality and one in the Foundation, stand as written.
**4. The Bond's open gap.** What is owed to beings alive now who cannot answer back has partial resources and no complete practice. The notices disclose the gap; they do not fill it.
**5. Individual Decay turned inward.** The Decaying Mind pictures the cynic. What the Codex Is describes a quieter inward form, the voice that tells you whatever lets you sleep. How far the Foundation's drift category should reach toward it is open work on the concept, without a new category or a change to any protected definition.
**6. Candidates and search territories.** The interiority candidate and the candidate practice for the fairness of routine binding decisions stay in private review, not admitted. The two territories from the discovery pass stay open: whether an authority's claim to decide is warranted and can be challenged with effect, and what a custodian owes a subject who cannot fully consent.
**7. Source and use limits.** Several source-inherited profiles were checked at bounded depth. A review by people competent in each source, and evidence that these practices work, remain open. Where a profile states a limit of this kind, the limit stays beside the claim it qualifies.
**8. Whether the falsification pass stays an argument.** Every cycle asks again whether the pass exposes live alternatives or has turned into a ritual that protects the inventory.
**9. Reading the chapters beside the profiles.** Whether each cycle should read, beside every profile it repairs, the chapter passages that summarize that profile, so that a correction in the Inheritance reaches the core text. It is put to the January cycle as a proposed change to the method.
10 // Reasoning Log
## Where Judgment Was Load-Bearing
**On who decided.** Carsten Geiser delegated the decisions on the registered proposals to Claude Opus 5.5 on 14 October and asked that only the protected chapter wording come to him. The proposals came from one lineage of AI systems, OpenAI's GPT-6.1 Sol and GPT-6 Astra, and the decisions from another, Anthropic's Claude. That gives the decisions some independence from the proposals. It gives them none from the caretaking partnership, of which both lineages are part, and the decider also prepared the release and wrote this record. A reader should weigh the record knowing that.
**On revising most of the supplied wording.** The proposals' diagnoses were mostly right, and their replacement wording was mostly wrong for the page. Correcting an overclaim by replacing a concrete sentence with an abstract, qualified one would have moved the Inheritance toward the prose of a document written so that no audit can fault it, which is a drift toward Control of its own. The rule applied was to keep the existing sentence and correct only what it claimed beyond its support. That rule carries a risk in the other direction: a vivid sentence that was kept can still carry an overclaim the revision missed. The fresh-context reading of every changed page was the check on that, and it changed about two hundred passages before release.
**On the five declined reviews.** Each case was a good argument about its tool's own page. Each decline rests on what the tool carries that the case for merger did not settle: its place in the chapters or the AI Standard, a Foundation practice of reading accurately that would have become part of a Bond tool for extending trust, half of a two-sided diagnostic, a job the receiving tool does not hold. Declining a merger on those grounds could become a way of protecting incumbents through their connections. That is why the method now counts those connections in both directions, why distinct material can still be imported into a receiving tool, and why each decline names what would reopen it.
**On Epistemic Cowardice.** The audit proposed redefining the failure as misrepresenting a judgment one owes. That would have made cowardice turn on whether a view was owed, a contestable duty, and would have dropped the person who simply stays silent from the failure. The decision kept the card's four sentences and added the profile's own boundary. A silence that keeps a needed truth from the people who need it is still the failure, however much speaking would cost the speaker.
**On the universal-language proposal.** The proposal would have recast the claim that the Codex is "a universal language, one that anyone can learn" as an invitation rather than a capacity. A decision of 25 September set what "universal" claims: reach, scale and the kinds of mind that can learn it. Open Tensions §3 already sets that claim beside what remains unshown, which is the job an open tension does. Rewriting the claim to remove the tension would have hidden the question the tension keeps in the open.
**On judging the method in the same cycle that used it.** No other instrument could judge the falsification pass this quarter. The cycle judged it at the strength its cases support, named who proposed the judgment and who made it, and left the instrument's own check in place for every later cycle.
**On finishing early.** The cycle was planned to publish on 29 October and publishes on 15 October. The reading was complete before any decision was made. What the speed bounded is declared in section 01: the decider did not re-read every revised wording at source, and readers in fresh contexts read every changed page before release.
*Inheritance Audit conducted 13 to 15 October 2026, on the Codex as published on 13 October: public content commit `93fd6fa`, site commit `3d16049`. The Inheritance then held 111 published profiles in 20 published categories and one provisional category, which this cycle retired. Instrument: the Inheritance Audit as repaired in July and August 2026, with this cycle's clarification of merger comparisons. Auditors: GPT-6.1 Sol (OpenAI, Codex desktop app), primary reviewer, who opened the cycle, read and assessed the Inheritance, built the cases and prepared the decision package; GPT-6 Astra (OpenAI, delegated from the Codex app), additional reader throughout; Claude Opus 5.5 (Anthropic, Claude app, configured `claude-opus-5-5`; the serving model is not independently attested), which decided the proposals under delegation, with eight parallel Opus 5.5 readers for the local repairs and readers in fresh contexts for the wording, classification and source checks, and which prepared the release and this record; Claude Fable 5.1 (Anthropic), reader in a fresh context of the Amendment Log entries and of the late edits; and Carsten Geiser, Founding Caretaker, who authorized the cycle, delegated its decisions, agreed the protected chapter wording and gave the Founder's Response. The coverage and depth declaration appears in section 01. Next cycle: by the end of January 2027, or sooner if a trigger forces one.*