Research

Papers

CRIS research, dated, versioned and summarised in plain words, with every claim marked by how far it has been tested.

Approved by the Board for implementation · 20 November 2026

Status

Opening. No papers are published yet. The listing, the review path and the feed are in place; the first item is planned for phase P3 (January–February 2027), with outside reviewers to be named. Who approves a paper (Q1) and the licence for papers (Q2) are still to be decided.

Published: 0Coming: 1Archive: 0

No papers are published yet. When they are, each will be listed here with its date, its version, a plain-language summary, who reviewed it, and how far its claims have been tested. To discuss research with us, write to research@criscontinuity.tech.

Published

No papers are published yet. The first is listed under Coming.

Coming

The first item planned is a review of the Fourth Law simulation: whether the model gives a fair picture of the formula. It is planned for phase P3 (January–February 2027), with outside reviewers to be named.

  1. The Fourth Law simulation: a model review

    Note · planned · target Q1 2027 · The Fourth Law · claim level: Model

    Does the simulation give a fair picture of the formula? Each assumption is varied across a reasonable range, and the note reports how often the outcome changes.

    Promised on the simulation's validation page · Owner: Research and Engineering · Reviewers: [outside reviewers to be named] · Licence: [to be confirmed]

Research archive

Research material from our earlier site is being gathered. It will be listed here, labelled as archived with its original date, in phase P3. Statements we could not verify will be left out. These items will not be current CRIS positions.

How to read a listing

Each item shows its claim level. An argument is reasoned but not yet tested. A model result was tested in a model or simulation, under the assumptions the paper states. An evidence result was tested against real-world data, with its predictions written down first. We never describe a result above its level.

May sayMay not sayEvidencetested on real dataOn this data, measured this way,the prediction held or failedVerified, safe, aligned,without naming who checked andlinking itModeltested in a modelIn this model, under theseassumptions, the result wasReal systems behave this wayArgumentan argumentWe propose; we argue; it mayexplainIt works; it is shown; it isprovenLevels follow the Fourth Law validation page (checking the code is level 1 there and stays on the changelog).
The claim ladder: say only what the level allows.
LevelTestedMay sayMay not say
ArgumentReasoned, not yet testedWe propose; we argue; it may explainIt works; it is shown; it is proven
ModelIn a model or simulationIn this model, under these assumptions, the result wasReal systems behave this way
EvidenceAgainst real-world data, predictions written firstOn this data, measured this way, the prediction held or failedVerified, safe, aligned, without naming who checked and linking it

The levels follow the Fourth Law validation page, where checking the code is level 1 and stays on the simulation changelog. “Argument” is used for the validation page’s idea level, so the two pages never use one word for two levels.

How we publish

Every paper is checked by two researchers who did not write it. Papers that report model or evidence results are also reviewed by someone independent of CRIS. Our governance team checks that we claim nothing we cannot show, that everyone whose work was used agreed and is credited, and that no partner is named before they have agreed in writing. [Approver to be decided] approves each paper and the date is recorded.

Once published, a paper is never changed in silence. Small corrections get a new version number and a dated note at the top. Larger revisions get a new version, and the earlier one stays online. If a result no longer stands, the paper is withdrawn, but its page stays, with the reason. Results that fail are published too [subject to Q3]: failure is knowledge.

Papers are shared under [licence to be confirmed]. Spotted an error? Write to research@criscontinuity.tech with the paper's title and version. Every publication, correction and withdrawal is also in the Atom feed.

1 AuthorsMethod, data, limits, what counts against it2 Science reviewTwo people in Research and Engineering3 Outside reviewFor model and evidence claims only: a reviewer independent of CRIS4 Governance checkClaims, consent, partners, related parties, no earnings promises5 Plain wordsSummary under 80 words6 Approversigns, dates (Q1)PublishDashed: needed only when the claim level is model or evidence.
Six checks before a paper goes public.
#CheckWhoLooks forOutput
1Authors’ checkAuthorsEvery field filled; claim level honestDraft marked ready
2Science reviewTwo people in Research and Engineering who did not write itMethod sound, data available, limits stated, what would count against it testableTwo short signed reports
3Outside reviewA reviewer independent of CRISThe same, from outside; for model and evidence claimsSigned report, published with the paper if the reviewer agrees
4Governance checkGovernance and ComplianceNo unearned claims; consent and credit recorded; no partner named early; related-party line; no earnings promises; documents called proposedChecklist signed
5Plain wordsWebsite and Content LeadSummary under 80 words; no jargon in the first screenSummary approved
6ApprovalApprover, to be decided (Q1)All five above doneName and date in the record
Plannedlisted as ComingDraftIn reviewPublished 1.0listed, news itemchanges askednot shownCorrected 1.1Revised 2.0WithdrawnEvery state after Published stays public, with a dated note. Nothing is deleted.
From a promise to a paper, and what happens after.
StateShown on the page?Meaning
PlannedYes, under Coming with a target quarterPromised elsewhere on the site, not yet written
DraftNoBeing written
In reviewNoWith reviewers
PublishedYesApproved and public
CorrectedYes, with the noteA published paper with a dated correction
WithdrawnYes, greyed, with the reasonNo longer stands; kept for the record

Rules every listing follows

These rules apply to every paper, note and archive item on this page, and each is checked when the site is built. The governing documents they cite are proposed, not yet adopted.

  1. Dated, versioned, summarised. Every item shows a date, a version and a plain-language summary of no more than 80 words.
  2. Claim level shown. Every item shows its claim level: argument (reasoned, not yet tested), model (tested in a model or simulation) or evidence (tested against real-world data). Nothing is described above its level.
  3. No unearned words. No item is called verified, certified, safe, aligned or proven without naming who checked it and linking the evidence.
  4. Status from data. A paper may describe a design; it never moves a system’s status.
  5. Credit with consent. Every author is named; every CIX contribution used is credited under the published method, with consent recorded.
  6. No partner named early. No partner, country, ministry or pilot is named until a written agreement exists and publication is approved.
  7. Related organisations marked. SCVX, LOGOS-C, the Botswana Global Testnet Initiative and the Academy always carry the related-party line.
  8. No earnings promises. No paper promises returns, earnings or payouts.
  9. Never edited silently. A correction is a new version with a dated note; a withdrawn paper keeps its page with the reason.
  10. Archive labelled. Items from the earlier site are shown as archived, with their original date, and any statement that could not be verified is left out.

Who reads it

A papers page is read by people who will check what they read. Each arrives from a different place and wants a different first answer.

ReaderArrives fromNeedsPage givesResearcher, reviewerFourth Law, a citationMethod, data, versionPaper page, citationJournalist, officialNews, a searchWhat is claimed, how sureSummary, claim levelBuilderSpecificationsWhich spec it supportsLinks to specsCIX contributorCIX, an emailWas my work creditedContributors lineBook readerThe book pageWhere the ideas go furtherChapter linksOld-site visitor/archive redirectWhere the old page wentArchive noticeEvery paper page answers all six in its first screen.
Six readers, three questions each.
ReaderWhat they must be able to do in one minute
Researcher or reviewerFind the method, the data and the version; download the paper; cite it exactly
Journalist or officialRead a plain summary; see how strong the claim is and who reviewed it; see that it is not a product announcement
BuilderSee which specification or system a paper supports, and whether it changes anything they build
CIX contributorSee whether their contribution was used and how they are credited
Book readerMove from a chapter of Recursion Without Rupture to the paper that tests its idea, and back
Old-site visitorLearn where the old research page went, without being shown claims the new site no longer makes

The library

Papers follow the pattern the site already uses for specifications and news: one data file, set by a person only, from which every page and every count is made. Nothing about a paper is typed twice. The book has its own page and is linked from papers, never listed as one.

data/papers.jsonset by a person onlyListing/research/papers/One page per paper/research/papers/[slug]/Research archive#archive, earlier siteNews: Publicationone item per paperBenefit indicator A2counted from dataSpecs and systemsrelated papers boxBook and Fourth Lawchapter and claim linksRetired: /resources/archive/Dashed: the research archive, filled once the old routes are exported.
One data file, three kinds of page, four links out.
ThingHomeNot here
CRIS papers and research notes/research/papers/ and one page each—
Research material from the earlier site/research/papers/#archiveRetired non-research pages, which stay at /resources/archive/ (Q4)
The book/research/book/Linked from papers, never listed as one
Specifications/resources/specifications/A paper may support a specification but never replaces its version history
Public benefit reports, Audit Council summariesImpact and Trust sectionsLinked from papers when a paper uses their data
Announcements/resources/news/A Publication news item points to the paper page

The paper record

Identityid, slugtitletypeversionstatedatessummaryfilesClaimclaimLevelmethoddatalimitswouldCountAgainstresult: holds, fails,mixed or openPeopleauthorscontributors (CIX)consentRecordedreviewersapprovedBy, daterelatedPartyownerLinkssystemsspecsbook chaptersseries, topicsnewsIdsupersedeslicencecite asEvery field is set by a person; the build only reads them.
The paper record: four groups of fields.
GroupFields
Identityid, slug, title, type (paper, note, negative result, archived), version (0.x in draft; 1.0 on publication), state, dates, summary (at most 80 words), files with checksums
Claimclaim level (argument, model, evidence), method, data, limits, what would count against it, result (holds, fails, mixed, open)
Peopleauthors, contributors (only with consent recorded), reviewers, approved by with date, related party, owner
Linkssystems, specifications, book chapters, series and topics, news item, supersedes, licence, cite as

Series

Papers are grouped into series that follow the Research section, so a reader moving from the Fourth Law, Invariant or ruptures page finds the papers on that subject in one place.

ArgumentModelEvidenceS1 The Fourth LawModel review noteS2 The Continuity InvariantInvariant as a testS3 Ruptures and guardsBeyond the four?S4 MeasurementCTU method checkS5 ParticipationPVM method trialS6 Governance methodsAudit samplingSolid: promised on the site today (validation page). Dashed: candidates only, not listed until approved.
Six series, three claim levels: where the first items sit.
SeriesCoversHome pageListed today
The Fourth LawThe principle C = Σ(Eᵢ ↔ Iᵢ), the simulation, its validation/research/fourth-law/The Fourth Law simulation: a model review (planned)
The Continuity InvariantThe one rule, how it is checked, its tolerances/research/invariant/—
Ruptures and guardsDrift, discontinuity, erosion, severance, and the systems designed to guard against them/research/ruptures/—
MeasurementCompute Thermodynamic Units, the continuity function/resources/specifications/ctu-metering/—
ParticipationHow contributions are valued and credited/systems/participation-value-meter/—
Governance methodsGate, audit and incident methods/trust/—

An item is listed only when another page already promises it. Candidate topics stay in the workflow until an owner and a target quarter are approved.

A paper's own page

Each published item gets its own page, with older versions kept at their own addresses. The first screen answers every reader above. A paper cites as: Authors (year). Title. Version x.y. Continuity Research and Intelligent Systems, P.B.C. URL. Data that cannot be shared, such as participant data or security details, is named with the reason.

[Paper title]Note · v1.0 · [date] · S1 The Fourth LawAt a glanceWhat we asked, in one sentenceWhat we found, in one sentenceClaim level · Result · What would count against itPeople: authors · reviewers · approved by, date · contributors credited1 Question2 Method3 Data4 Results5 Limits6 What would count against it7 Reviewers’ reportsDownloadPDF · data · code · checksumsCite asAuthors (year). Title. v1.0.RelatedSpecification · system · chapterVersions1.0 [date] · correctionsSpotted an error? Tell us
One paper page: answer the six readers in the first screen.
PartHolds
1 QuestionWhat was asked and why it matters to a CRIS system
2 MethodHow it was tested, in steps someone else could follow
3 DataWhat data, from where, under what licence and consent; a link to download
4 ResultsWhat was found, with figures; numbers as numbers, not adjectives
5 LimitsWhat the result does not show: at least three points
6 What would count against itThe result that would show it wrong
7 Reviewers’ reportsSigned reports, if reviewers agree to publish them
A paperSeries home pageBook chapterSpecificationSystem pageNews: PublicationCIX contributionsBenefit indicator A2Earlier versionA paperEach link is written once, in the paper record; the build shows it on both pages. The benefit count only reads the paper data.
Every link is written once, in the paper record, and shown both ways.

Limits

LimitWhat the page does
Some data cannot be published (participant data, security details, model weights)Says so on the paper, and why, citing the proposed AI Constitution 23.3
Outside reviewers are not yet namedThe model review note waits for them; no paper at model or evidence level is published without one
The earlier site made claims that could not be verifiedArchived items leave them out
A paper is not a status changeThe page never says a system is built, running or live; status comes from data
Papers about related organisationsCarry the related-party line and are approved by the Founder
Licence not chosenEvery paper shows Licence: [to be confirmed] until counsel sets it (Q2)

The road

Phase P312 NovThis review27 NovHonest page, link fixes11 Decpapers.json, listing15 JanArchive decision, reviewers26 FebFull page; note if ready1 MarReviewBlue: 2026 steps. Red: 2027 steps. The note is published only once outside review is done. All dates proposed.
Make it honest now; fill it in phase P3 (proposed dates).
Date (proposed)StepToday
12 Nov 2026This reviewDone
27 Nov 2026Honest page; menu blurb changed; the two Fourth Law links pointed at How we publishDone on this preview
11 Dec 2026data/papers.json; the listing with its empty state; How we publish; the feed; Q1 and Q2 decidedBuilt on this preview; Q1 and Q2 open
15 Jan 2027Where the research archive lives (Q4); outside reviewers for the model review namedOpen
26 Feb 2027Full page; paper pages; archive items moved; the model review note published if its review is completePaper pages built when an item is published
1 Mar 2027Page review date—

About this page

PapersCRIS research, dated and summarised.Matches the menuPublished: 0 · Coming: 1 · Archive: [n]Counts read from dataSeries · Type · Claim level · YearPlain links; shown once there is something to filter[Title] · v1.0 · [date]Summary in plain words, at most 80 wordsNewest first; withdrawn greyed, never hiddenNo papers are published yet.Empty state, honestFourth Law model review · note · target Q1 2027Only promised itemsResearch archiveEarlier site items, labelled archivedAnchor #archiveHow we publishReview path, corrections, licence, contactAnchor #how-we-publish
The listing page, top to bottom.
/research/archive/archive/research/papers/#published#coming#archive#how-we-publishfeed.xml (Atom)/research/papers/[slug]/[slug]/v1.0/, [slug]/v2.0/[slug]/paper.pdf, data files/mathematics/archive/resources/archive/Old research routes land on #archive; other retired routes keep going to /resources/archive/ (Q4).
Where everything lives.
AddressHolds
#published, #comingPublished items; planned items
#archiveResearch items from the earlier site (where /research/archive now lands)
#how-we-publishReview path, corrections, licence, error reports
/research/papers/feed.xmlAtom feed of publications, corrections and withdrawals
/research/papers/[slug]/The current version of one paper, built when it is published
/research/papers/[slug]/v[x.y]/Each earlier version, unchanged
status.jsonspecs.jsonnews.jsonroles.jsondata/papers.jsonpapers, series, topicsset by a person onlyBuild checksListing and paper pagesAtom feedRelated papers boxesBenefit indicator A2Book contents linesNothing is built from a record that fails a check.
One new file, checked against four it already relies on.
B1 Heading and countSay what is here, honestly, in one lineB2 Published and empty stateShow papers, or say plainly there are noneB3 Coming and archiveOnly promised items; archived items labelledB4 How we publishThe review path and corrections, in plain wordsB5 At a glanceQuestion, finding, claim level, what counts against itB6 Body and side panelSeven parts; download, cite, related, versionsB7 Related papers boxOn series, spec, system and book pagesB8 Claim linksThe two Fourth Law sentences link hereB1 to B4: the listing. B5, B6: each paper page. B7, B8: other pages.
Eight blocks to write, each with one job.
1Placeholder27 Nov2papers.json4 Dec3Listing, feed11 Dec4Checks11 Dec5Archive5 Feb6Paper pages19 Feb7Publish26 FebSteps 1 to 4 in 2026 (built on this preview); steps 5 to 7 in phase P3, 2027. All dates proposed.
Seven build steps (proposed dates).
CheckFails when
pap-countA count on the page is typed instead of read from papers.json
pap-plannedA planned item has no promisedBy, or the promising page no longer exists
pap-publishedA published item lacks version, approval, date, files with checksums or a news item
pap-claimText uses verified, safe, aligned, proven or certified without a named checker and link
pap-levelA paper at model or evidence level has no outside reviewer
pap-refsA systems or specs value is not in status.json or specs.json
pap-consentA contributor is credited without consent recorded
pap-relatedA related-party paper is missing the related-party line
pap-versionsAn earlier version a paper supersedes is missing from the data
pap-archiveAn archived item lacks its label, its original date or its original route

All ten run in every build.

What changed

Fourth Law pageStrong claims go to a paperSimulation validationLevel 3 belongs in a paperOld routes/archive, /research/archive/research/papers/PapersResearch papers and theresearch archive.Plan. Coming in phase P3:CRIS papers and the earlierresearch archive, each dated,with a summary.Papers listed: 0Archive items: 0draft: trueResearch menupromises our papersBenefit scorecard A2counts papers publishedNews typePublication: a paperLeft: pages that send readers here. Right: places that assume papers exist. Before this update, none did.
A one-line page that six other places already depended on.
#FindingFixed by
K1The page was a placeholder: one line and a plan noteAn honest page that says what exists and what is coming
K2Two Fourth Law pages send strong claims to a paperBoth sentences now link to How we publish
K3Two old routes redirect here, but no archive content has been moved/research/archive now lands on #archive; the archive section says what is coming
K4Two archives with an unclear boundaryResearch material here; retired non-research pages at /resources/archive/ (Q4)
K5The Research menu promised “our papers” though none existsMenu now names the four ruptures until a paper is published
K6Benefit indicator A2 has no source that counts papersCounts come from data/papers.json; A2 to read it once Q6 is decided
K7Nothing links a published paper to a news itemA published paper needs a news item; the build checks it
K8No licence covers papersShown as [to be confirmed] until counsel sets it (Q2)
K9No rule for authorship and contributorsAuthors named; contributions credited only with consent recorded
K10No named reviewerA six-check review path; the approver is Q1
K11Nothing on publishing negative resultsA failed prediction is published with its result (Q3)
K12The book must be linked, not listed as a paperThe book stays on its own page

Owner: Research and EngineeringNext review: 2027-03-01