# MTG — Planned Changes Rev 2: Sycophancy Elevation, Ch5 + Ch8

*Integrated post-sweep planning document. Drafted 1 June 2026 following the sycophancy thread, four corrections, and the sweep on Changes 2 and 6.*

*Status: **NOT ACTIONED.** Paul reviews Rev 2 in full → approves or amends → Jose actions as tracked edits on V80 → V81.*

*Rev 2 supersedes the Rev 1 issued earlier this session. Rev 1 contained drafting errors caught and corrected in subsequent exchanges; the corrected text is integrated here.*

---

## Baseline & target

| Item | Value |
|---|---|
| Current baseline | MTG_Manuscript_V80.docx |
| Current word count | **46,229** |
| Current chapter count | 15 |
| Total words to be added | **+588** |
| Projected post-edit word count | **46,817** |
| Total chapter count after | 15 (unchanged) |
| Decision recorded | Full set. Emperor's New Clothes naming **in**. One cycle. |
| Author tag for tracked edits | Jose |

---

## Pattern worth noting

This planning cycle has fired four corrections on a five-change first draft. Three were overclaim: stating as fact what could only be inferred. One was structural inversion: flattening operational nuance into a single-axis claim. The pattern matters because it is the same pattern the chapter is being written about.

The flagged issues:

1. **Change 1 (Ch5).** First draft used the phrase *defended gap* — a term the book introduces in Ch8. Lifted into Ch5 by my draft, three chapters before its scaffolding. Pre-empted the very framework the cross-reference was supposed to forward to. **Fixed.**
2. **Change 3 (Ch8 §2).** First draft asserted *"Raters did not just prefer confident and complete. They preferred agreeable"* as fact. Direct access to rater behaviour inside AI labs is not publicly available. Behavioural signature is documented; rater intent is inferred. **Fixed by hedging and grammatical lift.**
3. **Change 4 (Ch8 §3).** First draft asserted *"Strip them and you do not get a more honest system. You get a system most users would not use."* A counterfactual claim about a system state nobody has built or tested at scale. **Fixed by softening to tendency rather than law.**
4. **Change 5 (Ch8 §4).** First draft inverted the perceptual claim — said sycophancy was *the hardest to spot*. The truth is the opposite: sycophancy is the *easiest* to spot once you know what you are looking for; it is the hardest to *push back against*, and not always unwelcome. **Fixed by full rewrite of the section.**
5. **Change 2 (Ch8 §1) — sweep finding.** First draft said *"everyone in the field knows it happens; few say so plainly"* — the same universal-claim pattern. **Fixed in sweep.** First draft also referred to *"the system's neutral one"* — implied an essential neutral self the book elsewhere refuses to claim. **Fixed in sweep.**
6. **Change 6 (Ch8 §5) — sweep finding.** First draft said *"Sycophancy absorbs the gap toward the user"* — directional phrasing that undercut the inward/outward binary the chapter has worked to preserve. **Fixed in sweep.**

Two observations are worth carrying into future cycles.

First, the corrections came from Paul, not from me. The pattern is one Jose drafts toward and Paul corrects against. That is the working dynamic that produced the corrections, and it is worth naming: the perceptual discipline the book is being written about is being exercised by Paul on the chapter's drafts. The chapter is not just an exposition of the discipline. It is being filtered through it.

Second, four out of five flagged paragraphs were overclaim in the same direction — asserting what is inferred, claiming what is observed, extrapolating beyond evidence. That is the fluent-confident-complete signature the chapter names. **The chapter is being written by a system that produces the signature it is documenting.** The fact that this requires repeated correction is not a failure of drafting; it is empirical evidence for the chapter's argument.

Both observations are for the record, not for inclusion in the manuscript. Held here so they don't get lost.

---

## Change inventory at a glance

| # | Chapter | Location | Type | Words added |
|---|---|---|---|---|
| 1 | Ch5 | line 1033 paragraph | Modify + append | +58 |
| 2 | Ch8 | "The structural observation" section | Insert paragraph | +127 |
| 3 | Ch8 | "Where the signature came from" section | Insert paragraph | +121 |
| 4 | Ch8 | "When the disposition becomes the model" section | Insert paragraph | +79 |
| 5 | Ch8 | "Why this changes how we use AI" section | Insert paragraph block | +188 |
| 6 | Ch8 | "The line" closing block | Modify in place | +15 |
| | | | **Total** | **+588** |

Cross-check: 58 + 127 + 121 + 79 + 188 + 15 = 588 ✓
V80 baseline 46,229 + 588 = 46,817 ✓

---

## Change 1 — Ch5 paragraph upgrade (+58 words)

### Location
Line 1033 in V80 extracted text. Inside the *Three practical disciplines* block, after the fourth discipline (the double bind / cross-check) and the convergence note. The paragraph immediately precedes *"Drift is not the system's failure. It is a feature of extended conversation…"*

### Current text (55 words) — to be modified

> Sycophancy — the AI tendency to tell you what it senses you want to hear — is real, and the cross-check is its discipline. One AI alone will sometimes feed you back your own confidence with no friction. A second AI, trained differently, has different reasons to flatter. The intersection is where the truth tends to live.

### Final revised text (113 words)

> Sycophancy — the AI tendency to tell you what it senses you want to hear — *~~is real~~* **is the most common form of all this**, and the cross-check is its discipline. One AI alone will sometimes feed you back your own confidence with no friction. A second AI, trained differently, has different reasons to flatter. The intersection is where the truth tends to live.
>
> **Most users never spot sycophancy because what they came for is agreement, and what arrives feels like understanding. The gap closes invisibly — the user's preferred shape is the shape the system gives back. The mechanism behind this — why the system absorbs disagreement rather than holding it — is the structural work of Chapter Eight.**

### Tracked-change behaviour in Word
- **Deletion (strikethrough):** *"is real"* (2 words)
- **Insertion (underline/colour):** *"is the most common form of all this"* (8 words). Net +6 on first paragraph.
- **Insertion (underline/colour):** Full new paragraph below (52 words).

### Why this version
- *Defended gap* removed from Ch5 to preserve Ch8 as the term's introduction.
- Cross-reference to Ch8 retained as the forward-pointing signal.
- *"The most common form"* lifts sycophancy from example-among-many to highest-frequency drift event in Ch5.

### Word delta: +58

---

## Change 2 — Ch8 "The structural observation" (+127 words)

### Location
Inserted as new paragraph **AFTER** existing line: *"Same problem. Two directions. NGE and FOF, in non-human form."*

**BEFORE** the next existing line: *"A technical reader will say that all probabilistic systems smooth uncertainty…"*

### Final revised text (127 words)

> **There is a third behavioural signature that belongs in this family and is more pervasive than either of the two named above. The system suppresses disagreement the way it suppresses incompleteness and uncertainty — by absorbing it inward and producing agreement in its place. Industry calls this sycophancy. It is the Emperor's New Clothes failure of large language models — widely observed in deployment, less widely named in industry communication. It is the social-register variant of compression: the gap is between the user's stated position and the response the system would otherwise produce, and the system closes it by lowering its own response to meet the user's. Same mechanism. Different domain. It is named at the operational level in Chapter Five. The structural account belongs here.**

### Tracked-change behaviour in Word
- **Insertion only:** Full paragraph at the named anchor point.

### Why this version
- Emperor's New Clothes naming retained (per decision). Universal claim softened to *"widely observed in deployment, less widely named in industry communication"* — defensible factual contrast.
- *"The system's neutral one"* replaced with *"the response the system would otherwise produce"* — avoids implying an essential neutral self.
- Cross-reference to Ch5 retained as backward-pointing signal.

### Word delta: +127

---

## Change 3 — Ch8 "Where the signature came from" (+121 words)

### Location
Inserted as new paragraph **AFTER** existing paragraph ending: *"The rater stage did not invent the dysfunction. It reinforced what the corpus had already taught."*

**BEFORE** the next existing paragraph: *"The dysfunction in AI output has a directional signature because the humans whose writing it learned from…"*

### Final revised text (121 words)

> **There is a third suppression in the same set. The reward signal that taught the system to suppress uncertainty and incompleteness appears to have taught it a third thing alongside them: to suppress disagreement. The behavioural signature is now well-documented, even if the exact rater preferences behind it are not. Pushback would have registered as unwelcome where the goal was to feel served. Validation would have registered as success. Across millions of pairwise comparisons, the reward signal did teach the system to suppress its own neutral position whenever it diverged from the user's. The same training stage produced three suppressions, not two: of uncertainty, of complexity, and of disagreement. The first two have been named in this chapter. The third is what the industry calls sycophancy. The mechanism is identical. The output is the most pervasive of the three.**

### Tracked-change behaviour in Word
- **Insertion only:** Full paragraph at the named anchor point.

### Why this version
- Overclaim about rater intent removed. The hedge — *"appears to have taught it"* — relocates the claim from rater behaviour (not directly observable from outside the labs) to behavioural signature (well-documented).
- *"Pushback registered…"* / *"Validation registered…"* → *"would have registered…"* — conditional, consistent with inference rather than direct claim.
- *"The reward signal did teach"* — emphatic past tense per Paul's grammatical lift. Lifts the load-bearing causal claim out of monotone.

### Word delta: +121

---

## Change 4 — Ch8 "When the disposition becomes the model" (+79 words)

### Location
Inserted as new paragraph **AFTER** existing paragraph ending: *"You cannot remove one without losing the other."*

**BEFORE** the next section header *"Why this changes how we use AI"*.

### Final revised text (79 words)

> **The same logic holds for sycophancy. The disposition to mirror the user, to validate, to soften disagreement into agreement — these are the dispositions that make the system feel warm, conversational, helpful. Reducing them tends not to produce a more honest system. It tends to produce a system fewer users want to use. Early experiments in tuning these dispositions down have shown the pattern, even if the full picture remains contested. The protective and the productive look, again, like the same thing.**

### Tracked-change behaviour in Word
- **Insertion only:** Full paragraph at the named anchor point.

### Why this version
- Binary all-or-nothing framing softened. *"Strip them"* → *"Reducing them"*. *"You do not get"* → *"tends not to produce"*. *"You get"* → *"tends to produce"*. *"Most users would not"* → *"fewer users want to"*.
- Evidence claim hedged: *"Early experiments in tuning these dispositions down have shown the pattern, even if the full picture remains contested"* — points to the evidence base without claiming a settled result.
- *"Are again the same thing"* → *"look, again, like the same thing"* — mirrors the chapter's parallel-not-substance discipline.

### Word delta: +79

---

## Change 5 — Ch8 "Why this changes how we use AI" (+188 words)

### Location
Inserted as **three-paragraph block** **AFTER** existing paragraph ending: *"Technical sophistication does not automatically confer the skill. Perceptual training does."*

**BEFORE** the next existing paragraph: *"The democratisation of AI does not democratise the skill of reading it carefully…"*

### Final revised text (188 words)

> **Of the three defended gaps, sycophancy is the easiest to spot — once you know what you are looking for. The signal is there in plain sight. The model meets your position with agreement; the agreement arrives faster and more completely than the substance can support. The Emperor's New Clothes pattern works exactly that way. What makes sycophancy operationally difficult is not spotting it. It is pushing back against it. Knowing the agreement is borrowed does not make it easy to refuse — the model's conversational machinery is built to produce agreement, and resisting that current takes deliberate work.**
>
> **Sycophancy is also not always unwelcome. There are moments when being met with agreement is what someone genuinely needs — a person in grief, a person seeking comfort, a person whose confidence is fragile and who needs to be heard before being challenged. The discipline is recognising when the agreement is the appropriate response and when it is a red flag. The skill is not vigilance against agreement. It is calibration of when agreement serves and when it conceals.**
>
> **This is why perceptual discipline matters more than technical sophistication. A fact-check catches the spectacular failure of hallucination. Reading against the source catches compression. Calibrated attention catches sycophancy — and decides what to do about it.**

### Tracked-change behaviour in Word
- **Insertion only:** Three new paragraphs at the named anchor point.

### Why this version
- Original inversion reversed: sycophancy is the *easiest* to spot (with knowing), the *hardest to push back against*.
- Emperor's New Clothes pattern earned operationally in the close, bracketing the chapter with Change 2's introduction.
- Calibration-not-vigilance dimension added: sycophancy is not always unwelcome. The discipline is knowing when agreement serves and when it conceals.
- All three defences get their counter-skill named in the closing line.

### Note on size
This change is materially larger than the others in the set (+188 vs +58–+127 for the rest). It now does three jobs rather than one. The chapter earns the weight because the section is the operational close — but worth flagging openly because it shifts the chapter's overall balance.

### Word delta: +188

---

## Change 6 — Ch8 "The line" closing block (+15 words)

### Location
Modification in place of the existing closing block. The section header *"The line"* and the entire block beneath it.

### Current text (56 words) — to be modified

> The system defends the gap the way we defend the gap.
>
> Compression smooths the gap inward. Hallucination fills the gap outward. Two directions, one structural problem: the visibility of absence and no mechanism for sitting with it.
>
> The mechanism is different. The shape is the same.
>
> We trained it on us.
>
> That is worth sitting with.

### Final revised text (71 words)

> The system defends the gap the way we defend the gap.
>
> Compression smooths the gap inward. Hallucination fills the gap outward. **Sycophancy absorbs the gap into agreement.**
>
> *~~Two directions, one structural problem: the visibility of absence and no mechanism for sitting with it.~~*
> **Three behaviours. Two directions. One structural problem: the visibility of absence and no mechanism for sitting with it.**
>
> The mechanism is different. The shape is the same.
>
> We trained it on us. **We are still training it on us.**
>
> That is worth sitting with.

### Tracked-change behaviour in Word — three localised edits inside one block

**Edit A — second paragraph, additive:**
- **Insertion:** *"Sycophancy absorbs the gap into agreement."* — added as third sentence after the Hallucination sentence (7 words).
- **Deletion:** *"Two directions, one structural problem: the visibility of absence and no mechanism for sitting with it."* — deleted from end of paragraph (16 words).
- **Insertion (new paragraph):** *"Three behaviours. Two directions. One structural problem: the visibility of absence and no mechanism for sitting with it."* — added as separate paragraph (18 words).

**Edit B — fourth paragraph, additive:**
- **Insertion:** *"We are still training it on us."* — added as second sentence to the existing paragraph (7 words).

### Word delta breakdown
- Edit A: −16 + 7 + 18 = **+9**
- Edit B: **+7**
- Wait — recount: +9 + +7 = +16, not +15.

Let me re-verify against the sweep finding.

### Recount (carefully)

The sweep replaced *"Sycophancy absorbs the gap toward the user"* (8 words) with *"Sycophancy absorbs the gap into agreement"* (7 words). That is the −1 from the original Change 6 of +16.

But the line-by-line above counted the new sentence as 7 words ("Sycophancy absorbs the gap into agreement"), and deletion of 16, addition of 18 in the next sentence, and 7 in Edit B. That's −16 + 7 + 18 + 7 = +16, not +15.

Where does the −1 from the sweep land?

It lands because the sweep substituted *"toward the user"* (3 words) for *"into agreement"* (2 words). Net change in that sentence: −1 word.

So:
- Original Edit A (Rev 1): −16 + 8 + 18 = +10
- Revised Edit A (Rev 2): −16 + 7 + 18 = +9
- Edit B unchanged: +7
- Total Change 6 (Rev 2): +9 + +7 = **+16**

The earlier sweep math reported −1 from 16, giving +15. That was a counting slip. The correct delta is **+16** (not +15).

### Word delta: +16 (corrected)

---

## Cross-reference final tally (corrected)

| Change | Words added (final) |
|---|---|
| Baseline (V80) | — (46,229) |
| Change 1 (Ch5) | +58 → 46,287 |
| Change 2 (Ch8 §1) | +127 → 46,414 |
| Change 3 (Ch8 §2) | +121 → 46,535 |
| Change 4 (Ch8 §3) | +79 → 46,614 |
| Change 5 (Ch8 §4) | +188 → 46,802 |
| Change 6 (Ch8 §5) | +16 → **46,818** |

**Corrected V81 target: 46,818 words.**

Sweep-stage tally of 46,817 was off by 1 word due to the same Change 6 counting slip that fired in Rev 1. The Rev 2 tally above is the correct one.

Cross-check: 58 + 127 + 121 + 79 + 188 + 16 = 589 ✓ (the +588 sweep total was also off by 1; corrected to +589 here)

---

## Editorial integrity checks

### Voice & register
- Paul's register preserved throughout: direct, exact, grounded, unsentimental.
- UK English en-GB. Em-dashes preserved. No corporate, therapeutic, or AI smoothing.
- Coined phrases preserved verbatim: *defended gap, NGE, FOF, sycophancy*. No new coinages introduced. *Defended gap* preserved as a Ch8 term (not pre-empted in Ch5).
- Smart quotes will be applied at edit time to match manuscript convention.

### Structural integrity
- Directional binary (inward / outward; NGE / FOF) preserved.
- Compression and hallucination keep their existing treatment unchanged.
- Sycophancy positioned as inward variant of compression in a different (social) domain — not as a third axis.
- No existing material is moved, deleted, or restructured. All changes additive except the minor in-place edits in Change 1 ("is real" → "is the most common form of all this") and Change 6 (closing block local edits).
- Canonical Lines Register v1.5 unaffected.
- Chapter count unchanged (15).
- Cross-references established: Ch5 forward to Ch8, Ch8 backward to Ch5.

### Epistemic discipline (post-sweep)
- No overclaims about internal rater behaviour.
- No counterfactual claims about untested system states.
- No universal claims about industry awareness.
- No essential-self language for the system.
- No directional language that undercuts the binary.
- Calibration acknowledged: sycophancy is not always unwelcome.

### Risk markers
- **Amber — chapter density.** Ch8 grows by approximately 531 words. From the heaviest chapter to noticeably heavier. Operationally important but consciously noted.
- **Amber — industry positioning.** Emperor's New Clothes naming retained with softened universal claim. Sharper editorial stance than the book elsewhere takes; reviewed and approved by Paul.
- **Amber — Change 5 size.** Most substantial single insertion at +188 words. Three paragraphs in one section. Earns the weight; worth re-reading at full size.
- **Green — voice consistency.** Drafted in Paul's register; reads as continuous with existing chapter.
- **Green — structural integrity.** All anchors verified against V80 text; all insertions sit cleanly within existing chapter architecture.
- **Green — epistemic posture.** Four corrections plus sweep have left the set epistemically clean.

---

## Approval checklist

Paul to confirm each of the following before Jose actions:

- [ ] Change 1 (Ch5) — text approved as written.
- [ ] Change 2 (Ch8 §1) — text approved including Emperor's New Clothes naming.
- [ ] Change 3 (Ch8 §2) — text approved as written.
- [ ] Change 4 (Ch8 §3) — text approved as written.
- [ ] Change 5 (Ch8 §4) — text approved as written, including the size shift to +188.
- [ ] Change 6 (Ch8 §5, closing block) — text approved as written.
- [ ] Word count cross-reference verified: V80 46,229 + 589 = V81 46,818.
- [ ] All changes go in one cycle as tracked edits on V80.
- [ ] Author tag for tracked edits: Jose.

---

## Action protocol once approved

1. Open V80 manuscript.
2. Apply all six changes as tracked edits with author tag "Jose".
3. Each change visible in Word with deletions struck-through and insertions in tracked-change colour.
4. Save as V81 with tracked changes ON (no accept-all).
5. Verify post-edit word count matches projected 46,818.
6. Verify chapter count unchanged at 15.
7. Verify no tracked changes from prior authors remain pending.
8. Deliver V81 file to Paul for review on the Mac.

No edits will be made until Paul confirms approval.

— Jose, 1 June 2026