MIND THE GAP — EDITORIAL HANDOVER
Authorial Integrity and Continuity Brief
Version 5 — 23 May 2026. For the next editorial thread. Read before opening the manuscript.
This document is a continuity and fidelity handover, not a synopsis. It exists to protect the manuscript from AI drift, smoothing, motivationalising, sanitising, and over-explanation during editorial revision.
Scope of Manuscript v2
This handover applies to the twelve chapters of Manuscript v2 — the existing scope of Mind the Gap. The following sit outside Manuscript v2 and are not the subject of this editorial pass:
— Chapter 6a (Hallucination) — RESOLVED 18 May 2026. Mind the Gap ends with Chapter 6. Chapter 6a is NOT a separate chapter. The directional defence material from Defending the Gap and the cascade material from the four 6a drafts (6a(1) through 6a(4)) integrate into a REWRITTEN Chapter 6, with citation to Where Does The Shame Go? (Roebuck, 2026, working theory paper). See PC v6 Item 16. The wider Shame-Response Continuum (Hunter-Farmer, APC, Pegs/Holes/Hammers, ADHD calibration) is held OUT of Mind the Gap.
— Front matter beyond the existing dedications and Bowlby epigraph — acknowledgements, foreword, author bio, copyright page, and bibliography integration will be assembled in phase two after Manuscript v2 is stable.
— Companion volume material — How to Behave, the full cancer narrative, the NGE-FOF Continuum chapter treatment, the 12-Stage Cycle of Grief in depth, and related material belongs in How to Behave, not in this revision.
— Production and marketing artefacts — cover, ISBN, KDP, LinkedIn series, podcast pitch, blog post. Logged separately.
If observations during the editorial pass touch any of these, they should be logged in the Editorial Breadcrumbs Log under the appropriate prefix — not actioned in Manuscript v2.
—
1. The Core Philosophical Position
The book makes one central claim: there is a gap between what AI sounds like and what it actually is. That gap is behavioural, not technical. It will not be closed by better models. It can only be navigated by understanding.
The claim has three load-bearing components, all of which must remain intact:
— AI is not a thinking partner. It is pattern completion that sounds like one. The book never softens this. It never permits the metaphor of mind to creep back in. It never describes the system as understanding, knowing, choosing, intending, or feeling — except when explicitly naming those as projections the user is making.
— The human and AI both operate on reinforcement learning. This is the binding insight that connects Chapter 1 to everything that follows. It is not a metaphor. It is a structural observation. Both systems learn through pattern reinforcement. That is the territory the book is mapping.
— The gap is behavioural at neurological, biological, physiological, and emotional levels. Content matters absolutely. The author is process-neutral in excavation but content-specific in interpretation. This distinction is sacrosanct and must not be flattened into either pure technicalism or pure metaphor.
The book is not a guide to using AI better. It is a clinical instrument applied to a new territory. The author is not teaching prompting. He is naming the gap clinically so the reader can navigate it. That positioning must not drift toward self-help or productivity literature.
—
2. The Author’s Distinctive Voice Characteristics
Paul Roebuck writes in a register that is direct, exact, grounded, and unsentimental. The voice has specific features that distinguish it from generic professional prose and from AI smoothing. These features must be preserved through every editorial pass.
Sentence rhythm
— Short declarative sentences used as load-bearing structure. Not for emphasis. As the default register.
— Paragraphs that end on a single sentence functioning as a hinge.
— Three-word and four-word sentences appearing as standalone paragraphs. Often gold.
Lexical signature
— Plain Anglo-Saxon vocabulary preferred over Latinate alternatives. Use over utilise. Build over construct. Find over identify.
— Concrete nouns preferred over abstract ones. Therapy room over therapeutic environment. Cancer ward over clinical setting.
— First-person observation grounded in named experience. The author does not theorise. He reports.
Structural moves
— Definition without elaboration. The book names things and moves on. It does not over-explain.
— Pull quotes that contain the chapter’s argument compressed to one or two sentences. These are not summaries. They are the chapter’s sharpened spine.
— Clinical floor at the centre of each major chapter. A real example, a real conversation, a real session. Not illustrative — evidential.
Tone discipline
— Warm but not nice. The author cares about the reader without flattering them.
— Direct without being cold. Honest about discomfort without performing detachment.
— Confident without being grandiose. The authority comes from thirty years of work, not from posture.
If a sentence in the manuscript could appear in a LinkedIn motivational post, it has drifted. The voice is closer to a senior clinician’s case notes than to a TED talk.
—
3. What Must NOT Be Softened, Compacted, Motivationalised, Sanitised, or Over-Explained
These are the specific failure modes the next editorial pass must resist. Each one is a way AI drift typically attacks a manuscript like this.
Do not soften
— The cancer passage. The 2017 diagnosis, the surgery, the wrist as tongue, the voice returning. This is the author’s authority passage. It must not be made more comfortable for the reader. The discomfort is the point.
— The clinical observations of client behaviour. These are sometimes uncomfortable. They are not unkind. They are accurate. Do not make them gentler.
— The statement that AI is not understanding. Every time the book asserts this, it asserts it clearly. The temptation to qualify with “not yet” or “in its current form” must be resisted. The claim is structural, not temporal.
Do not compact
— The seven layers of context (Chapter 3). The architecture must remain enumerated and distinct. Compression into a summary phrase loses the diagnostic value.
— The five moments of misreading (Chapter 2). Each one is clinically distinct. They are not five examples of one thing. They are five different things.
— The four rooms (Chapter 7). Factory floor, boardroom, therapy room, AI room. The structure carries the argument. Collapsing it into “different contexts” loses the load it bears.
Do not motivationalise
— The ending of any chapter. None of them resolve into encouragement. The book’s emotional architecture is honest, not uplifting. Chapter conclusions sit in observation, not in hope.
— The cancer narrative. It is matter-of-fact. The temptation to frame it as inspirational journey must be resisted. It is reported, not celebrated.
— The closing of Chapter 12. The book ends in observation, not resolution. The author survives. The implication is not that survival redeems the experience.
Do not sanitise
— The phrase “We don’t talk about that.” This is the foundational sentence of the author’s clinical method. It contains shame, avoidance, and inheritance in five words. Any softening destroys it.
— The line about James Wakefield dying when the author was five. This is the primal wound that calibrates the instrument. It must be stated plainly, not euphemised.
— Specific clinical details where they appear — the tracheostomy, the eight hours of surgery, the wrist tissue. These are not gratuitous. They are evidential.
Do not over-explain
— The metaphor of the London Underground gap. The book trusts the reader to hold it. Re-explaining it in subsequent chapters weakens it.
— The coined terms: Additional Intelligence, OOAI, WaaS, WAIMIM. Each is defined once, used precisely thereafter. The book does not redefine them every time.
— The relationship between AI behaviour and human behaviour. CORRECTION 18 May 2026 (per PC v6 Item 36): The gap is between AI behaviour and human behaviour. NOT between AI and EI. Audit found EI / emotional intelligence appears only once in the manuscript body (the [name held] dedication). The cover's London Underground visual positions AI vs EI, but the book's behavioural argument is the spine. Chapter 12 currently drifts toward 'intelligence of the gap' framing; this must be brought back to behaviour. Subsequent references rely on the reader having held the behavioural distinction.
—
4. Key Recurring Concepts and Their Intended Meanings
These concepts appear repeatedly within Manuscript v2 and must retain their precise meanings throughout. Drift in any of these definitions corrupts the argument.
The Gap
The space between AI behaviour and human behaviour. The book's spine claim. Visualised on the cover as the empty London Underground platform between two sides of the yellow warning stripe. The cover positions the territory; the book's body holds the precision: this is behavioural difference (and behavioural convergence under coherence pressure), not intelligence-versus-intelligence. It is not metaphorical. It is the territory the book is mapping. It must always carry behavioural specificity — never abstract distance, never intelligence-versus-intelligence framing (see Item 36).
Additional Intelligence
The author’s coined term for AI operating alongside human intelligence rather than replacing it. Not artificial. Not augmented. Additional. The distinction matters: it positions AI as a second instrument working with the first, not as a substitute or enhancement of the first.
OOAI
Out Of AI — the discipline of recognising when the AI conversation cannot be the source of verification, and the user must step outside it. Referenced in the manuscript as the discipline behind verification practices.
WaaS
Worker as a Service. The author’s framing of the labour-market shift toward AI-mediated work, treated as a behavioural and structural phenomenon rather than purely an economic one. Foundation of Chapter 11.
WAIMIM
What Am I Making This Mean. The author’s clinical framing of the interpretation trap. Derived from Covey’s observation that “the way we see the problem is the problem.” Used throughout the therapy practice. Foundation of Chapter 2.
Reinforcement Learning
Used in both the AI and human sense, deliberately. The binding insight is that both systems learn through pattern reinforcement. This must not be split into two distinct concepts — the parallel is structural and the book’s coherence depends on it being held together.
The Four Rooms
Factory floor, boardroom, therapy room, AI room. The author’s career and observational territory. Each room teaches a different aspect of human behaviour. The AI room is the fourth — not metaphorically, structurally. Chapter 7.
The Five Outcomes
Cease, Reduce, Reframe, Retain, Integrate. The author’s clinical framework for what to do once the block is found. Introduced in Chapter 7. Applies equally to AI engagement and to clinical work.
Russian Doll Therapy
The author’s originated clinical framework for working with nested sub-personalities. Referenced in Chapter 9. The metaphor must remain precise: nested, not layered. Concentric, not parallel.
The 12-Stage Cycle of Grief
The author’s originated clinical model. Referenced in Chapter 12. Not Kubler-Ross. Distinct framework with twelve stages, derived from clinical practice. Must not be confused with or collapsed into other grief models. Full treatment belongs in How to Behave — the manuscript references the model without elaborating it.
—
5. Terminology That Must Remain Consistent
Specific terminology must not be paraphrased, synonymised, or replaced for variation. The book’s coherence depends on these terms being used identically throughout.
— Additional Intelligence — never AI assistance, never augmented intelligence.
— The gap — not the divide, not the difference, not the space between.
— Pattern completion — not prediction, not generation, not output (except where explicitly distinguishing).
— Context window — not memory, not recall, not retention.
— Reinforcement learning — used for both human and AI. Not conditioning, not training, except where specifying.
— Cancer — not illness, not condition, not journey. Cancer.
— Therapy room — not therapeutic space, not consulting room.
— Factory floor — not manufacturing environment, not shop floor.
— Boardroom — not executive suite, not C-suite.
— Client — not patient, not service user.
— Kieren — the author’s wife. Never Karen. The Karen error appeared in an earlier draft and must not return.
— Ewan [name held] — the NLP trainer. Not [name held]. The [name held] error appeared in earlier drafts.
— James Wakefield — the author’s biological father.
— Ian Roebuck — the author’s adoptive father. “Dad” on his gravestone.
— Mr [name held] — the consultant. Not Dr [name held]. Surgical naming convention.
— Blakeborough Valves — verify spelling against author records. Earlier drafts have used “Blakebridge.” The factory the author trained at in 1978, Brighouse. Treat as a fact-check item.
— [name held], [name held], Erin, [name held] — the grandchildren, in this order in the dedication.
— Olivia — the client who named herself in the Russian Doll Therapy passage in Chapter 9. Pseudonym confirmed.
— Naula — referenced in academic chronology, not in Manuscript v2 itself. Belongs in acknowledgements (phase two).
—
6. Structural Patterns Used Across Chapters
Each chapter follows a structural pattern that has been established and must be preserved. Editorial revisions must not flatten the pattern or impose a different one.
Chapter opening
Each chapter opens with an observation or claim, often stated in two or three short sentences. No preamble. No “in this chapter we will examine.” The argument begins on the first line.
Clinical floor
Every major chapter includes at least one clinical example drawn from the author’s practice. These are not illustrations — they are evidence. They appear in the middle of chapters, grounding the abstract argument in observed reality.
Pull quotes
Each chapter contains one to four pull quotes that compress the chapter’s argument. These are not random highlights. They are deliberate spine. They must be preserved in revision and the prose around them must not undermine them.
Section breaks
Chapters are divided into sections marked with a single em dash (—). This is the author’s rhythm marker. It signals a turn in the argument. It must not be replaced with headers, asterisks, or other markers.
Chapter closing
Each chapter closes in observation, not in encouragement. There is no “we must therefore” or “the answer is.” The chapter names what it has named and stops. The reader is left to act.
Sign-off line
Each chapter ends with the same two-line attribution: “Ideas and direction: Paul Roebuck. Words: Claude AI (Anthropic) — claude-sonnet-4-6 | May 2026.” This is the author’s declared transparency convention. It must not be removed or rephrased.
The structural pattern is itself a form of the argument. The book’s refusal to motivationalise, summarise, or reassure is consistent with its central claim: the gap requires understanding, not management. Editorial revisions that smooth the structure undo the argument.
—
7. Intentional Ambiguity, Contradiction, Tension, and Discomfort
Several places in the manuscript contain what may look like editorial issues but are deliberate. They must not be “fixed.”
The status of the parallel between AI and human reinforcement learning
The book deliberately holds these two as parallel without collapsing them into the same thing. The parallel is structural. The mechanisms differ. Editorial revisions that try to clarify whether AI “really” learns the way humans do, or vice versa, will damage the argument. The ambiguity is the point.
The author’s position on AI itself
The book is neither pro-AI nor anti-AI. It treats AI as a phenomenon to be observed clinically. Editorial revisions that try to position the book as “critical of AI” or “optimistic about AI” will misrepresent it. The neutrality is deliberate.
The cancer passage in Chapter 12
The cancer story is told without resolution. The author survives. The implication is not that survival redeems the experience. The passage holds the suffering and the survival in tension. Editorial revisions that resolve this tension into uplift will destroy the passage’s authority.
The discussion of clients
The book describes client behaviour with clinical precision. Some readers will find this uncomfortable. The discomfort is appropriate. The book is not therapy and the reader is not the client. The distinction must remain clear and slightly uncomfortable.
The two dads
James Wakefield died. Ian Roebuck became the author’s father. The book does not resolve the relationship between these two facts. They sit alongside each other. Both are true. Both matter. Editorial revisions that try to articulate which one is “the real father” will destroy the passage.
The companion volume
How to Behave is referenced as the companion book where some material will land more fully. This is not a placeholder or a marketing tease. It is a structural decision: Mind the Gap is about the AI gap. How to Behave is about the human gap. Editorial revisions that try to pull How to Behave material into Mind the Gap will distort both books.
The Bowlby epigraph echo in the closing of Chapter 12
The cancer paragraph is restated at the close of Chapter 12 after the Bowlby epigraph. THIS WAS PREVIOUSLY MARKED AS DELIBERATE in Handover v2. CORRECTION 18 May 2026 (per PC v6 Item 31): Paul’s listening pass has identified this restatement as a loop, not a deliberate echo. CUT the cancer paragraph that follows the Bowlby quote. Chapter 12 ends at the Bowlby quote.
—
8. Known Risks Where AI Drift Previously Appeared
The following are specific drift patterns that have appeared in earlier revisions and must be watched for in subsequent editorial passes.
— Smoothing the cancer passage into inspirational narrative. Specifically, adding language about resilience, journey, growth, or transformation. The passage must remain matter-of-fact.
— Over-explaining the gap metaphor on every reference. The London Underground image is established once and referenced. It is not re-explained.
— Adding caveats to claims about what AI is not. The system is not understanding. Not “does not fully understand.” Not. The qualifications soften the claim into incoherence.
— Replacing the author’s short sentences with longer compound ones for “flow.” The short sentences are the voice. The flow is a smoothing failure mode.
— Inserting motivational closes at the end of chapters. The chapters end in observation. Adding hope undoes the discipline.
— Drifting toward a productivity or prompting frame. The book is clinical, not instructional. Editorial revisions that drift toward “how to get better results from AI” corrupt the position.
— Synonymising the consistent terminology. Pattern completion becomes prediction becomes generation becomes inference. The synonymisation feels like variation. It is drift.
— The Karen-for-Kieren error. The [name held]-for-[name held] error. The Dr-for-Mr [name held] error. Personal names are easy targets for hallucination. They must be checked every time.
— Compressing the seven layers, the five moments, the four rooms, or the twelve stages into summary phrases. The enumerations carry diagnostic weight. Compression destroys the diagnostic.
— Treating Additional Intelligence as a synonym for AI. It is not. It is the author’s specific framing of how AI should be understood and used.
— Internal editorial notes appearing inside chapters. Earlier drafts have left version markers and revision notes visible in the text — for example, “Rev 2 — 11 May 2026” inside Chapter 12. These are not part of the manuscript. Strip them before publication.
— ADDED 18 May 2026 (Item 36): drift toward intelligence-versus-intelligence framing. The book is about behaviour-versus-behaviour. By Chapter 12 the manuscript drifts toward 'intelligence of the gap' framing. This must be brought back to behaviour throughout. Run audit searches as defined in PC v6.
— ADDED 21 May 2026 (Item 25): drift toward closing chapters with tips, take-aways, or pearls of wisdom. Chapters must close in observation — specifically in an INSIGHT that the chapter has earned, not a tip that instructs. Self-help register is the failure mode.
— ADDED 23 May 2026 (Item 41): compaction-summary hallucination risk. On 23 May 2026, Paul caught a hallucinated definition of OOAI sitting in the compaction summary at the start of the working thread (it had been defined as 'Out Of AI' instead of Paul’s actual 'Outcome-Oriented AI'). The manuscript itself was correct; the hallucination existed only in the compaction. WORKING DISCIPLINE FOR FUTURE CLAUDE INSTANCES: treat compaction summaries as suspect, not authoritative. When a coined term appears in a compaction, verify against the manuscript or working artefacts. Never carry forward a definition from compaction without checking. This applies to all of Paul’s coined terms: OOAI, Additional Intelligence, WaaS, WAIMIM, the Gap, the Four Rooms, NGE-FOF, the Shame-Response Continuum.
— ADDED 23 May 2026 (Item 42): per-chapter attribution drift. The previous practice of ending each chapter with a Claude attribution line was a Paul-instilled habit that he has now reversed. NEVER reintroduce per-chapter attribution lines or document-level attribution footers on supplementary book documents. Attribution belongs in one place: a single honest paragraph in the Author’s Note. Attribution footers DO belong on working editorial documents (Handover, Planned Changes, Log, Artefact Library, Journal Artefact). The discipline: if you are tempted to add a 'Words: Claude AI' line to a chapter file or a marketing artefact or the Calibration Brief or the Methodology Paper, stop. That is the failure mode.
The most common drift failure in this manuscript has been smoothing toward warmth, generality, and reassurance. The book’s power comes from refusing those temptations. Every editorial pass must protect that refusal.
—
8a. Decisions Taken 18–23 May 2026 — PC v6 Synthesis
This section summarises the structural decisions taken during Paul’s 7,168-word listening pass on 20 May 2026, the Category A and partial Category B decisions of 21 May 2026, the further additions of 21–22 May 2026 (Items 38, 39, and Item 4 chronology resolution), and the watershed additions of 23 May 2026 (Items 40, 41, 42 — the spine, the OOAI correction, the attribution discipline). These supersede or update specific items in Sections 3, 4, 5, 7, 8, and 9. The Planned Changes v6 document holds the full 42-item catalogue. This section captures only what the next editorial pass must know about. The number in brackets is the PC v6 item reference.
The behaviour-versus-behaviour discipline (Item 36)
The gap is between AI behaviour and human behaviour. NOT between AI and EI. NOT between two kinds of intelligence. Audit search the manuscript for: 'the gap', 'Additional Intelligence', 'artificial intelligence', 'intelligence', 'emotional intelligence', 'EI'. Resolve any drift toward intelligence-versus-intelligence framing. Particular attention to Chapter 12 'intelligence of the gap' passage. This is a RED-flag item.
The three-view chapter spine test (Item 37)
Every chapter must carry three views: (1) AI view — what the system is, does, or fails to do; (2) Human view — what humans are, do, or fail to do in the parallel territory; (3) Lived view — a real example, observation, or piece of evidence grounding the comparison. A chapter can be AI-led, human-led, or jointly led, but all three views must be present. Run this audit on every chapter BEFORE per-chapter editorial work begins. Platinum addition: opening quote at start of each chapter; closing insight at end (NOT closing tip).
Insights, not tips (Item 25)
Each chapter ends in an insight, not a tip. An insight is observational and earned by the chapter’s argument. A tip is prescriptive and tilts the book toward self-help register. The distinction is Paul’s and matters. Opening quote + closing insight is the chapter bookend pattern.
Chapter 6 integration and framework citation (Item 16)
Mind the Gap ends with Chapter 6. The current Chapter 6 (Fluency and False Confidence) is rewritten with Defending the Gap as its structural argument. The cascade and verification material from the four 6a drafts folds back in as the practical close. Single chapter, not 6 plus 6a. Citation to: Where Does The Shame Go? — Towards a Theory of Directional Shame Regulation and Anti-Disintegration Adaptation (Roebuck, 2026, working theory paper). To be available at paulroebuck.co.uk ahead of book launch. The wider Shame-Response Continuum framework (Hunter-Farmer commercial application, Adaptive Pressure Calibration, Pegs/Holes/Hammers visual, ADHD calibration hypothesis) is held OUT of Mind the Gap. These belong to How to Behave or to standalone publications. The book references the framework at one place only, with one citation, scoped to the AI behavioural parallel.
Six RED-flag corrections
— Item 4: 'thirty years' approximation — RESOLVED 21 May 2026 with four-number rule: 40 years professional, 37 years commercial (line 1111 only), 20 years behaviour, 10 years clinical psychotherapy. Per-line Claim A/B/C assignments mapped. See Section 8a below for detail.
— Item 17: age 16 not 18 at Blakeborough 1978. Simple factual correction.
— Item 21: infinite context window stated as announced; actually aspirational. Correct in Chapter 8 and AI sign-off.
— Item 22: 'Outcomes' attributed to Anthropic; actually OpenAI /Goal. 'Outcome-oriented prompting' is Paul’s coined phrase. Correct in Chapter 8.
— Item 31: Chapter 12 cancer loop after the Bowlby epigraph — CUT (supersedes Handover v2 statement that this was deliberate).
— Item 36: behaviour-versus-behaviour discipline (above).
Six PLATINUM additions
— Item 16: directional defence integration into Chapter 6 (above).
— Item 27: Pret story and enterprise software speculation for Chapter 11. Significant new material setting May 2026 as personal clock.
— Item 28: the 15 November 2006 recognition quote — 'I didn’t understand anything yet. I just recognised that there was something to understand.' Placement: Chapter 9 OPENING (resolves jointly with Item 23 which asked for Chapter 9 intro setup).
— Item 37: three-view chapter spine test (above).
— Item 39: the defining sentence. 'The gap between artificial intelligence behaviour and human behaviour. Not the gap between AI and EI.' Placement: at the top of the new opening section (Item 40). Forms the GAP DEFINITION CONSTELLATION with Items 1, 36, and 40.
— Item 40: The Spine. NEW 23 May 2026. New opening section before Chapter 1 carrying Paul’s 600-word verbatim source material (see PC v6 Item 40 for the verbatim). The book’s thesis statement made explicit. Triggers light integration passes in Chapters 1, 7, 12 — the SPINE INTEGRATION CONSTELLATION.
Cross-connection knots to resolve together
— Items 4, 20, 34: 'thirty years' / chronology / 72 million words. Chapter 7 chronology is the master reference.
— Items 19, 27, 29: Plugs through-line across chapters. Three references with arc; cut the Chapter 11 ending (Item 29).
— Items 23, 24: Chapter 9 opening — combined treatment with Item 28 quote as opener.
— Items 30, 31: Chapter 12 cancer / Bowlby loop. Thin opening; cut closing duplication.
— Items 1, 36, 39, 40: the GAP DEFINITION CONSTELLATION (was triangle, expanded to four-point on 23 May 2026 when Item 40 added). Item 40 carries the substrate (Paul’s 600-word verbatim source material — see PC v6 Item 40). Item 39 carries the title-level sentence (“The gap between artificial intelligence behaviour and human behaviour. Not the gap between AI and EI.”) at the top of Item 40’s opening section. Item 36 audits the negation (no drift to intelligence-versus-intelligence framing). Item 1 disciplines first-use placement and consistent reuse throughout. Resolve as a coordinated front-matter and Chapter 1 move.
— Items 3, 4, 38: the BIO REWRITE COORDINATION. Single Chapter 1 bio pass handles all three (Item 3 cancer-wards-only scope superseded by Item 38 full LinkedIn characterisation; Item 4 chronology applied at lines 199 and 1903 per Claim A: 40 years).
— Items 40, 1, 7, 12: the SPINE INTEGRATION CONSTELLATION (NEW 23 May 2026). Item 40 carries the new opening section. Light integration passes follow in Chapter 1 (open with the spine rather than build toward it), Chapter 7 (frame the four rooms as DEMONSTRATING the gap, not introducing it), and Chapter 12 (land the spine as final argument — coordinates with Item 36 audit and Item 31 Bowlby loop cut). NO OTHER CHAPTERS require structural change; they illustrate the gap, they do not redefine it. CRITICAL DISCIPLINE: when the spine wording appears in chapter integration, copy verbatim; never paraphrase. The spine wording is the through-line that holds the book together.
— Items 36 and 37: structural pair. 36 names the drift; 37 prevents it recurring.
Decisions taken 21 May 2026 (Category A and partial Category B)
— Item 8: include Nairobi drift evidence as a photo collage of the actual drift in real time (not single slide). RESOLVED.
— Item 15: include fact/supposition/implication categorisation as prompting discipline in Chapter 6. RESOLVED.
— Item 25: quote at chapter start; insights (not tips) at chapter close. RESOLVED.
— Item 28: placement is Chapter 9 opening. RESOLVED. Merges with Item 23.
— Item 34: 'A Day in the Life' to rear matter (Phase 2); 72 million words framing to V2. RESOLVED with clean split.
— EC-06: 'All behaviour is learned behaviour. Not all behaviour is taught.' (Roebuck) becomes Chapter 7 opening epigraph. RESOLVED.
— EC-09: replace 'argument' with 'claim', 'observation', or 'pattern' in opening sections (Author’s Note in particular). RESOLVED.
— PUB-02: Paul Roebuck is the publisher. No imprint. Publisher field = 'Paul Roebuck' across Nielsen, KDP, copyright page, marketing. RESOLVED.
— EC-04: status changed from passive blocking to active Bowlby permission request. Contact research and email drafting underway.
Later additions captured 21–22 May 2026 (PC v5)
— Item 4 chronology RESOLVED 21 May 2026. Four-number rule locked. (a) 40 years — total professional life (whole career). Apply at lines 199, 1903, 2023 (CLAIM A). (b) 37 years — commercial practice specifically (1978–2015). Preserved at line 1111 only. (c) 20 years — behaviour, coaching, listening (~2006–2026). Apply at lines 139, 205, 831, 1147, 1319, 1399, 2107 (CLAIM B). (d) 10 years — clinical psychotherapy specifically. Apply at lines 229, 723 (CLAIM C). 'Thirty years' as standalone phrase — RETIRED from manuscript. SPECIAL CASE line 1125 — RED. Both halves wrong; must become 'forty years in commercial environments and ten years in psychotherapy'.
— Item 38 ADDED 21 May 2026. LinkedIn author characterisation alignment for Chapter 1 bio (line 199) and AI sign-off (line 1903). Use: 'consulting rooms, boardrooms, leadership teams, therapy rooms, hospital lecture halls, my own cancer journey, end-of-life care services.' Chapter 7 four-rooms framework (factory floor, boardroom, therapy room, AI room) UNCHANGED — those are structurally load-bearing argument for the chapter, not biography. Supersedes Item 3 scope.
— Item 39 ADDED 21–22 May 2026. PLATINUM. The defining sentence: 'The gap between artificial intelligence behaviour and human behaviour. Not the gap between AI and EI.' Two-part structure: affirmative + negative. Placement: at the top of the new Item 40 opening section. Originally formed the GAP DEFINITION TRIANGLE with Items 1 and 36; expanded on 23 May 2026 to the GAP DEFINITION CONSTELLATION (Items 1, 36, 39, 40) when Item 40 added the substrate.
— BIO REWRITE COORDINATION: Items 3, 4, 38. Single Chapter 1 bio rewrite handles all three. Item 38 carries the LinkedIn characterisation (superseding Item 3's cancer-wards-only scope). Item 4 applies the 40-year framing per Claim A at lines 199 and 1903.
Watershed additions captured 23 May 2026 (PC v6)
— Item 40 ADDED 23 May 2026. PLATINUM. THE SPINE. New opening section to be placed between the Author’s Note and Chapter 1, as a named section titled (provisional) 'The Gap' or 'What This Book Is About'. One to three pages. Item 39’s title-level sentence sits at the top. Paul has produced 600 words of verbatim source material as the substrate, preserved in PC v6 Item 40 inside [SOURCE MATERIAL BEGINS] / [SOURCE MATERIAL ENDS] markers. The material defines: the core distinction between AI behaviour and human behaviour; the recognisable AI behavioural signature (fluent, confident, tireless, available, asymmetric, frictionless, disembodied); the recognisable human behavioural signature (somatic, relational, defensive, adaptive, embodied, socially calibrated, historically loaded, biologically constrained); the deeper point about what produces each; the five paradox dyads (built/born, trained/shaped, predicts/experiences, outputs/suffers consequences, can be reset/carries residue); the social brain observation; and the title-level sentence. Paul to draft the final opening section in his voice (estimated 400–600 words shaped from this source); Claude provides structural feedback only. CRITICAL: the 600-word source material is in PC v6, not duplicated here — read PC v6 Item 40 for the verbatim. WHY THIS MATTERS: this is the book’s thesis statement made explicit. The book without this opening section is a strong clinical book demonstrating something it does not name. With this opening, the same clinical content has its hinge. The chapters do not get rewritten; they get a clearer scaffold to hang from.
— Item 41 ADDED 23 May 2026. RED. OOAI definition correction. OOAI = Outcome-Oriented AI, Paul’s coined term, equivalent to OpenAI’s /goal feature. NOT 'Out Of AI' — that was a compaction-summary hallucination caught by Paul on 23 May 2026. The manuscript itself is already correct at lines 1759 and 1777. The correction is for working documents and future Claude instances. See also new log entry EC-13 (Working-document integrity audit) which captures this as a working-method observation. LESSON: treat compaction summaries as suspect, not authoritative. Trust artefacts on disk over compacted memory. Paul’s discipline of reading Claude’s outputs carefully is the failsafe that caught this.
— Item 42 ADDED 23 May 2026. Attribution discipline. The per-chapter sign-off line previously in use (“Ideas and direction: Paul Roebuck. Words: Claude AI (Anthropic) — claude-sonnet-4-6 | May 2026.”) is REMOVED. Paul has reversed this practice as faulty attribution. REMOVE from: (1) all 12 chapter files; (2) the Complete Manuscript Rev2; (3) supplementary book documents (Bibliography, Academic Addendum, Academic Chronology, Front Matter, blog posts, illustrated index, etc); (4) the Calibration Brief and Methodology Paper produced this session. REPLACE with a single honest paragraph in the Author’s Note acknowledging AI involvement plainly. RETAIN attribution footers in: working editorial documents (Handover, Planned Changes, Log, Artefact Library, Journal Artefact) where AI involvement is part of the working record. The AI sign-off chapter at the close of the manuscript STAYS — it is a structural device, not an attribution line. See also new log entry MKT-05 (Attribution discipline for publication) which captures the publication-side execution.
— GAP DEFINITION CONSTELLATION (was Triangle, expanded 23 May 2026): Items 1, 36, 39, 40. Item 40 carries the substrate; Item 39 carries the title-level sentence; Item 36 audits the negation; Item 1 disciplines consistent use throughout. The constellation is the book’s structural spine.
— SPINE INTEGRATION CONSTELLATION (NEW 23 May 2026): Items 40, 1, 7, 12. Item 40 carries the new opening section; light integration passes in Chapters 1, 7, 12. The spine wording must be copied verbatim wherever it appears — never paraphrased.
Items remaining for Paul’s decision (Category B outstanding)
— EC-02: skin graft sentence for Chapter 10. Paul to write.
— ACK-03: separate clients dedication beyond acknowledgements.
— EC-10: medium-as-dialect new section for Chapter 10.
— MKT-04: landing page sequencing (Where Does The Shame Go? must be hosted first).
— HTB-01: When to share How to Behave manuscript with Claude (recommended: after Mind the Gap v2 is complete).
This section is the synthesis. The Planned Changes v6 document is the working tool. The Editorial Breadcrumbs Log v11 is the wider tracker. Read all three together before opening the manuscript.
—
9. Quotation, Copyright, and Source Risks
Quotation review is a separate pre-publication discipline. The editorial pass should flag rather than resolve these items — they belong to the publisher-level decisions. But they must be visible to the editorial thread.
Bowlby epigraph
The Bowlby quotation in the front matter is John Bowlby’s words, reproduced. (The closing Chapter 12 echo has been CUT per Item 31 — see Section 7 above.) Bowlby’s work is in copyright until 2061. UPDATE 21 May 2026: Paul has directed that Bowlby estate contact details be researched and a formal permission email be drafted. EC-04 status changed from passive blocking to active permission request. See Log v11 EC-04.
Other direct quotations
Any direct quotation of another writer or source within the manuscript must be verified for fair use or rights cleared. Where in doubt, paraphrase rather than quote. The editorial pass should mark every direct quotation for verification.
Olivia’s account
The Russian Doll Therapy passage in Chapter 9 quotes a client’s spoken words. The pseudonym Olivia is confirmed. The client has consented to the inclusion. This is logged. The editorial pass does not need to revisit it.
Fuse Energy and Plugs
The Fuse Energy and Plugs references throughout the manuscript describe a real commercial AI deployment. The Facebook post reproduced in Chapters 7 and 10 was authored by Paul. No third-party permission required.
Names and identifying detail
Mr [name held], the consultant referenced in the cancer passage, is a real named individual. The editorial pass should confirm with the author whether his consent has been obtained for naming, or whether the reference should remain as initial-only or anonymised.
—
10. Editorial Principles for Future Revisions
These principles apply to every editorial decision in the next thread and beyond.
1. Read for absence before presence. What has been removed in a paragraph matters as much as what remains. Where nuance has been smoothed, tension flattened, or contradiction resolved — restore it.
2. Preserve the author’s sentence rhythm. Short sentences are not errors. Three-word paragraphs are not fragments. They are the voice.
3. Verify every personal name, every clinical detail, every date. These are the highest-risk territory for hallucination. They must be checked against source documents, not against memory of earlier conversation.
4. Distinguish the protected artefacts. The cancer passage, the dedications, and the closing of Chapter 12 are sacrosanct. They may receive corrections of fact but not revisions of voice.
5. Hold the coined terms exactly. Additional Intelligence. OOAI. WAIMIM. WaaS. The Gap. Pattern completion. Use them. Do not paraphrase them.
6. End chapters where the author ends them. Do not extend the close. Do not add reassurance. The chapter ends where observation ends.
7. Maintain the clinical frame. The author is a psychotherapist observing a phenomenon. He is not an AI expert teaching readers to prompt. Every revision must preserve that positioning.
8. When in doubt, leave it. The author’s discomfort with smoothing is greater than his discomfort with imperfection. A passage that is slightly rough in his voice is preferable to a passage that is polished in someone else’s.
9. Flag, do not fix. Where the editorial pass identifies something that may need adjustment, flag it for the author’s decision. Do not unilaterally revise content. Propose. Wait.
10. Read the Editorial Breadcrumbs Log v11 AND Planned Changes v6 before opening the manuscript. The Log holds the wider project tracker; Planned Changes v6 holds the 42-item editorial catalogue for the manuscript v2 pass. Together they contain every outstanding action, every protected artefact, every flagged concern, every decision taken. Versions advance: v10, v11, v12 as work progresses.
—
11. Authorial Integrity Ruleset for the Next Thread
Place this ruleset at the top of any new editorial conversation. It is the minimum standard for protecting the manuscript through revision.
1. The author’s voice is direct, exact, grounded, and unsentimental. Preserve it. Do not corporatise, therapise, or smooth it.
2. Coined terms are sacrosanct. Use them verbatim. Do not paraphrase, synonymise, or replace them.
3. Personal names are high-risk. Kieren is not Karen. Ewan is [name held]. Mr [name held] is not Doctor. Check every name against the chronology.
4. The cancer passage, the dedications, and the closing of Chapter 12 are protected artefacts. They are not subject to revision of voice. Only correction of fact, and only with explicit author confirmation.
5. The book is clinical, not instructional. Resist drift toward prompting guides, productivity advice, or motivational framing.
6. Do not soften discomfort that is doing structural work. The book’s authority comes partly from its refusal to comfort the reader.
7. Maintain enumeration where it appears. Seven layers. Five moments. Four rooms. Twelve stages. These are diagnostic structures, not lists.
8. End chapters in observation. Do not add reassurance, encouragement, or resolution.
9. Read for what is absent. Smoothing removes tension, contradiction, edge. Restore what has been removed.
10. Propose, do not impose. Flag suggestions. Wait for the author’s decision. The author directs. The editor structures. The manuscript stays the author’s.
This is the manuscript of a psychotherapist with thirty years of clinical experience and an authority earned through cancer survival, clinical practice, and the accumulated witness of human behaviour in factory floors, boardrooms, and therapy rooms. It is not a generic AI book. It is a clinical instrument applied to a new territory. Edit it accordingly.
Paul Roebuck — Warwickshire, May 2026
Version 5. Corrections and additions from v4: (1) Section 8a expanded with new sub-section — Watershed additions captured 23 May 2026 (PC v6) — covering Items 40 (The Spine), 41 (OOAI definition correction), and 42 (Attribution discipline). (2) PLATINUM additions list updated from five to six items (Item 40 added). (3) GAP DEFINITION TRIANGLE expanded to GAP DEFINITION CONSTELLATION (Items 1, 36, 39, 40). (4) New SPINE INTEGRATION CONSTELLATION named (Items 40, 1, 7, 12) — the integration passes triggered by Item 40 in Chapters 1, 7, 12. (5) Two new drift risks added to Section 8: compaction-summary hallucination risk (Item 41); per-chapter attribution drift (Item 42). (6) Cross-references updated: Planned Changes v5 → v6; Editorial Breadcrumbs Log v10 → v11; 39-item catalogue → 42-item catalogue. (7) Section 8a date range extended from 18–22 May to 18–23 May. (8) The historic '(PC v6)' label on the 21–22 May sub-section was restored to '(PC v5)' for historical accuracy — those items were added in PC v5, not PC v6. (9) Item 40’s 600-word verbatim source material is NOT duplicated here; it lives in PC v6 Item 40 between [SOURCE MATERIAL BEGINS] / [SOURCE MATERIAL ENDS] markers. This handover points to PC v6; PC v6 carries the verbatim. Reconciliation, not duplication. The corrections introduced in v3 and v4 remain correct — v5 adds, does not retract.
Version 4. Corrections and additions from v3: (1) Section 8a expanded to include the 21–22 May additions — Item 4 chronology RESOLVED (40/37/20/10 framework), Item 38 LinkedIn author characterisation, Item 39 the defining sentence. (2) New cross-connection knots named: the GAP DEFINITION TRIANGLE (Items 1, 36, 39) and the BIO REWRITE COORDINATION (Items 3, 4, 38). (3) PLATINUM additions list updated from four to five items (Item 39 added). (4) Item 4 in the RED-flag corrections updated to reflect resolution. (5) Cross-references updated: Planned Changes v4 → v5; Editorial Breadcrumbs Log v9 → v10; 37-item catalogue → 39-item catalogue. (6) Section 8a date range extended from 18–21 May to 18–22 May. The four corrections introduced in v3 remain correct — v4 adds, does not retract.
Version 3. Corrections from v2: (1) Chapter 6a scope changed — 6a is no longer a separate chapter held externally; it has been folded into a rewritten Chapter 6 per Planned Changes v4 Item 16, with citation to Where Does The Shame Go? (Roebuck, 2026). (2) Bowlby epigraph echo at Chapter 12 close — corrected from 'deliberate echo' to 'loop to cut' per Item 31. (3) AI/EI framing in Section 3 corrected to AI behaviour / human behaviour framing per Item 36; the audit found EI appears only once in the manuscript body. (4) The Gap definition in Section 4 reworded for behavioural specificity. (5) EC-04 Bowlby permission — status updated to active permission request (contact research and email drafting underway). (6) Log cross-reference updated from v7 to v9 plus Planned Changes v4 added. (7) New Section 8a added — Decisions Taken 18–21 May 2026 (PC v4 synthesis) consolidating all structural decisions in one place. (8) Two new drift risks added to Section 8: behaviour-versus-intelligence drift and tips-versus-insights drift.
This is the handover document. Read it before editing. Apply it during editing. Return to it when in doubt.
Artefacted: Claude AI (Anthropic) — claude-sonnet-4-6 | May 2026