b12 Process Learnings for b13+ Paper Prompts (2026m04d05)#

These learnings come from the b12 writing session (2026m04d05) and should be incorporated into all subsequent paper-writing prompts (b13–b17).

Lesson 1: Audience Assessment BEFORE Writing#

Problem: The b12 MMv1 was written as a single ~10K-word paper targeting “everyone.” This compressed critical concepts (PERFECT/PERFIDE, m2-to-m6 connection, th7 gates) to the point of making unjustified leaps.

Fix: Every paper-writing prompt MUST include a step where Claude assesses:

  1. Which distinct audiences the paper tries to serve

  2. Whether the requested word count gives sufficient space for each

  3. Per-section word estimates (what each section NEEDS vs. what it GETS)

  4. Whether audience-specific versions would serve readers better

Template text to add to prompts:

BEFORE WRITING: Assess the target audience(s). For each audience,
estimate per-section word counts. If the requested length is
insufficient to avoid leaps and gaps, say so and recommend
alternatives (longer single paper, audience-specific papers, or both).
Present your assessment for approval before writing.

Lesson 2: Reference Search IN Context#

Problem: The b12 MMv1 was written without the WoLC cross-traditional reference search. The search was done post-hoc and produced the session’s most valuable findings (Buddhist dependent origination, Erikson 8-stage parallel, Ashby’s Law grounding th4, Tuckman’s Storming = EQUAL).

Fix: Reference searches for cross-disciplinary connections should be done IN the same session as the paper writing, BEFORE the paper is finalized. The deep context from reading the axioms, theorems, and KB produces better structural matches than a cold-start search.

Template text to add to prompts:

REFERENCE SEARCH: Before finalizing the paper, search for independent
references from other disciplines that recognize fragments of the
model's structure. Use web search to check all references exist.
Report confidence level (CERTAIN/LIKELY/UNCERTAIN) for each.
DO NOT FABRICATE REFERENCES.

Lesson 3: Write All Audience Papers in One Session#

Problem: The MMv2 papers (5 audience-specific versions) were written in the same session that produced the MMv1 and the reference search. This was the correct decision: the deep context from the extraction KB, the axiom files, the reference search, and the MMv1 draft meant each audience paper could accurately capture the structural logic.

Fix: When a model spans multiple disciplines (as e7Day spans logic, theology, engineering, and psychology), plan for audience-specific papers from the start. Write them all in the session where the deep context is loaded.

Key audiences identified for e7Day (applicable to other models):

  1. Formal logicians / mathematicians

  2. Theologians / philosophers of religion

  3. Systems engineers / computer scientists

  4. Psychologists / social scientists

  5. General educated readers

Lesson 4: Adversarial Review BEFORE Refinement#

Problem: Without adversarial review, refinement risks polishing weaknesses rather than fixing them.

Fix: The b12 session produced adversarial review prompts for each audience paper. The workflow is:

  1. Write MMv2 drafts (done in one session with deep context)

  2. Run adversarial review on each (separate sessions, one per paper)

  3. Run refinement on each (separate sessions, incorporating review)

This three-phase workflow (draft → review → refine) should be standard for all papers.

Lesson 5: BEST Names Table#

Problem: The b12-math paper needed a symbol dictionary to bridge between mathematical notation and implementation-ready naming. This was not planned in advance but turned out to be one of the most useful artifacts.

Fix: Every paper with formal notation should include a BEST Names table (Brief, Explicit, Summarizing, Technical). Plan for it from the start.

Lesson 6: The Authorship Statement is Guarded#

Problem: The b11 authorship statement was updated and guarded by LLoL. The b12 papers initially used a slightly different version. All papers should use the SAME authorship statement.

Fix: Read the latest guarded authorship statement from the most recently reviewed paper (currently b11) and replicate it exactly. Do not paraphrase or modify the guarded text.

Lesson 7: File Naming Convention#

Use b-numbers (b11, b12, …), NOT a-numbers (a1, a2, …) for all paper references. The a-numbering was from an earlier convention. All references in text, filenames, and cross-links should use b-numbers.

Filename pattern: study_mmvN_YYYYmMMdDD_bNN-explicit-name.rst where N is the version, bNN is the paper number, and the explicit name describes the content.

Lesson 8: Downstream Prompt Updates#

The b13–b17 writing prompts should be updated to incorporate Lessons 1–7 above. A prompt for doing this update is below.

Prompt: Update Downstream Paper Prompts#

/effort max

TASK: Update the writing prompts for papers b13-b17 to incorporate
the process learnings from the b12 session.

Read:
1. .claude/CLAUDE.md
2. source/matheology/hell/ll/study/b/12/b12-process-learnings-for-b13.rst
   (this file --- the 8 lessons)
3. Each existing prompt:
   - source/matheology/hell/ll/study/b/13/b13-prompt-writing.rst
   - source/matheology/hell/ll/study/b/14/b14-prompt-writing.rst
   - source/matheology/hell/ll/study/b/15/b15-prompt-writing.rst
   - source/matheology/hell/ll/study/b/16/b16-prompt-writing.rst
   - source/matheology/hell/ll/study/b/17/b17-prompt-writing.rst

For EACH prompt, apply the following updates:
1. Add the "BEFORE WRITING: Assess audiences" template (Lesson 1)
2. Add the "REFERENCE SEARCH" template (Lesson 2)
3. Add instruction to write audience-specific papers if applicable (Lesson 3)
4. Add instruction to produce adversarial review and refinement prompts (Lesson 4)
5. Add instruction to include a BEST Names table for any paper with formal notation (Lesson 5)
6. Add instruction to read and replicate the latest guarded authorship statement (Lesson 6)
7. Replace all a-number references (a1, a2, ...) with b-numbers (b11, b12, ...) (Lesson 7)

Also update the extraction prompts (b13, b14) to reference the b12
extraction KB and extraction llog as methodological templates.

Save updated prompts in-place (overwrite). Create an llog entry
documenting what was changed.

IMPORTANT: Do NOT modify the content/substance of the prompts ---
only add the process improvements and fix the naming convention.
The models, axioms, and content instructions should remain as-is.