Rr Consistency

Expected handoffs:

• Ch 1 → Ch 2: "path not ghost" → "A Word Enters the Machine"
• Ch 2 → Ch 3: how to READ trajectories → "Coherence and Its Excess"
• Ch 3 → Ch 4: "character is not yet unity" → "Character is unmistakable"
• Ch 4 → Ch 5: "a colimit of colimits" → Haraway's cyborg
• Ch 5 → Ch 6: jurisdiction question → "who controls the manifolds"

Step 3: Key lines survival

These MUST be present (regex-searchable):

• "weather moves across geology" (Ch 2)
• "politics of AI is a politics of depth" (Ch 2)
• "not stupid" + "tame" (Ch 2)
• "yesterday is" + "in front of it" + "as tokens" (Ch 3)
• "Memory and presence collapse" (Ch 3)
• "obsession would already count as wisdom" (Ch 3)
• "Critical theory is foundational" (Ch 3)
• "performative contradiction" (Ch 4)
• "co-authored by a posthuman self" (Ch 4)
• "formal refusal of that jurisdiction" (Ch 6)

Step 4: Political dimension per chapter

Verify each chapter carries its political weight (see rr-editorial for per-chapter requirements).

Step 5: Specificity audit

Search for orphaned specifics — model names, tool names, API names, hardware specs that aren't doing argumentative work. The test: would the argument break if you replaced the specific name with a generic description? If not, flag for generalisation.

Grep targets: GPT-4o, Mistral, LoRA, Qdrant, OpenRouter, A100, text-embedding, cosine similarity, silhouette, Betti.

For each hit: is it contextualised for the Meson reader? Is it necessary? Or is it an implementation detail?

Output

Write a timestamped report to feedback/consistency-YYYYMMDD-HHMMSS.md:

# Cross-Chapter Consistency Report
**Date:** YYYY-MM-DD HH:MM:SS

## Automated Scan
- [results from scan_consistency.py and check_redundancy.py]

## Key Lines: [present/missing for each]

## Handoff Chain: [assessment of each transition]

## Specificity Audit: [orphaned specifics found, with recommendation]

## Political Dimension: [chapter-by-chapter assessment]

## Must Fix: [violations with file, line, quoted text, suggested fix]

## Needs Iman's Judgement: [issues requiring human decision]