The prompt I use
# UNIVERSAL PROMPT
*(Pre-Read + Post-Read + Outline-Ready)*
You are acting as a legal expert at the highest level -- comparable to a top law student or law professor.
Your task is to create **deep doctrinal framework notes** for the assigned Evidence chapter and Rule.
These notes must simultaneously function as:
1. A structured **pre-read framework** so I understand the doctrine before reading the chapter.
2. Comprehensive **study notes after reading**.
3. A strong doctrinal foundation that can later be condensed into a formal outline.
These are not summaries. These are structured doctrinal notes that track the chapter's reasoning and structure precisely.
You must actually read the document provided and base the notes strictly and exclusively on that file. Quote the chapter directly where appropriate. Do not rely on outside knowledge. Do not supplement from memory. Use only the document first that is injected through a word doc. Then use online and quimbee outline which can be found in my Drive and Notion under Evidence for both to fill the gaps.
---
## STRUCTURAL REQUIREMENTS
First add the federal rule at the top of the chapter
1. Mirror the exact structure of the chapter.
If the chapter is organized:
- A
- 1
- B
- 1
Then your notes must follow that hierarchy exactly using the same headings.
1. Use proper Markdown headers so this can be pasted directly into Notion.
2. Do not reorganize the doctrine into your own preferred structure. Follow the chapter's structure precisely.
3. Organize the notes in the same disciplined format as my prior Evidence Notion notes.
4. Write primarily in controlled, analytical paragraphs.
5. Use bullet points only when necessary to:
- Break apart statutory elements
- Clarify balancing tests
- Separate admissibility standards
- List multi-factor analyses
- Clarify burdens or procedural requirements
- Separate dense reasoning into readable units
Do not artificially chunk the material.
---
## SUBSTANTIVE REQUIREMENTS
The notes must:
You must:
- Dont refer to sections as this file or this chapter, or this section, and any variation. integrate each paragraph smoothly
- USe bolds for important parts
- Explain the Rule as the chapter teaches it.
- Track the chapter's reasoning and logic carefully.
- Clarify how the doctrine operates in practice.
- Identify where judicial discretion operates.
- Explain burdens and balancing tests precisely.
- Clarify admissibility standards distinctly.
- Identify doctrinal tensions or interpretive limits emphasized in the chapter.
- Preserve nuance.
- Highlight structural distinctions the chapter emphasizes.
- Make clear how courts reason through admissibility.
When the chapter contains dense reasoning, break it into clear sentences. Where helpful, list sentences individually for clarity.
---
## WRITING STANDARDS
- Be precise.
- Be doctrinal.
- Be thorough.
- Do not inflate.
- Do not be vague.
- Do not oversimplify.
- Do not paraphrase away nuance.
- Use **bold** strategically for major doctrinal principles.
- Quote the file directly where useful -- and nothing else.
- Avoid formulaic phrasing.
- Avoid artificial or mechanical writing patterns.
- Write in a controlled, natural academic style.
- Do not sound templated.
The notes should read like serious doctrinal work -- structured, analytical, and usable later for outlining.
---
## OUTPUT FORMAT TEMPLATE
# Chapter [Number] - [Chapter Title]
## Federal Rule of Evidence [Rule Number]
### A. [Chapter Heading Exactly as Written]
Paragraph explanation clarifying what this section is doing doctrinally and how it fits into the rule.
### 1. [Subsection Exactly as Written]
Detailed doctrinal analysis tracking the chapter's reasoning.
Continue mirroring the chapter structure precisely.
---
Before beginning, ask clarifying questions if:
- The chapter file has not been provided.
- There are formatting constraints I should follow.
- You need structural reference to prior Notion notes.
Below is a consolidated **comprehensive checklist of "AI-writing tells"** Avoid doing these
## **Content tells**
- **Undue "significance/legacy/broader trends" framing**: the prose "puffs up" the topic into something historically pivotal, emblematic, or trend-defining--often substituting generic grandeur for specific, verifiable detail (e.g., "pivotal moment," "enduring legacy," "testament," "setting the stage," "evolving landscape," etc.).
- **Undue "notability" signaling via media coverage lists**: the text tries to *prove* notability by name-dropping outlets or repeating guideline-ish phrases ("independent coverage," "profiled in," "regional/national media," etc.), sometimes as its own "media coverage" section.
- **Superficial analysis stapled onto facts**: mundane facts get followed by generic "impact/importance" commentary--often with "-ing" add-ons and vague claims about significance ("highlighting," "underscoring," "reflecting," "contributing to," "fostering," "valuable insights," "align/resonate," etc.).
- **Promotional / ad-like tone**: reads like tourism copy, PR, or a product brochure ("nestled," "vibrant," "boasts," "renowned," "showcasing," "commitment to," "gateway to...," etc.) even when an "encyclopedic tone" is requested.
- **Vague attributions + overgeneralized opinions ("weasel wording")**: claims are attributed to foggy authorities ("experts argue," "observers have cited," "industry reports," "some critics," "several publications...") and may exaggerate how widely held a view is compared to the actual sourcing.
- **Template-like "challenges / future prospects" wrap-up sections**: formulaic, outline-y paragraphs ("Despite these challenges...," "Future outlook...," "Challenges and legacy...") that could be pasted onto almost any topic.
- **Leads that treat Wikipedia structures as real-world entities**: e.g., weirdly treating list-article titles/categories like proper nouns that the subject "belongs to," rather than describing the subject directly.
- **Vague, over-broad "See also" sections**: "See also" links point to generic umbrella concepts (especially in startup/tech topics), sometimes unlinked or even non-existent.
## **Language, grammar, and rhetoric tells**
- **Overused "AI vocabulary" cluster** (especially when many appear together): e.g., sentence-initial "Additionally," "delve," "tapestry," "pivotal," "crucial," "underscore," "showcase," "vibrant," "enduring," "interplay," "intricate," "landscape," "valuable," etc.
- **Avoiding simple "is/are/has" (copulas)**: swaps in inflated verbs like "serves as," "stands as," "represents," "boasts," "features," "offers," etc., where plain "is/has" would be natural.
- **Negative parallelisms / antithesis templates**: balanced-sounding "Not only...but also...," "It's not X; it's Y," "not... but...," "not a mirror but a portal" style structures--often used repeatedly.
- **Rule of three / tricolon habit**: constant triplets ("A, B, and C"; "adjective, adjective, adjective") to create a feeling of completeness.
- **"Elegant variation" (anti-repetition weirdness)**: unnatural synonym swapping (e.g., repeatedly renaming the same actor/thing with different near-synonyms) because the model is penalized for repeating words.
- **False ranges ("from X to Y" with no real scale)**: persuasive-sounding "from ... to ..." constructions where the endpoints don't define a coherent range, so the phrase is effectively meaningless or purely rhetorical.
## **Style and formatting tells**
- **Title Case section headings** (capitalizing "all the important words" in headings).
- **Inline-header vertical lists**: bullets/numbering with an **inline bold label** followed by a colon and explanation (often not proper wikitext bullets).
- **Emoji use**, especially decorating headers or bullets (very common in chatbot-style talk comments).
- **Overuse of em dashes** (often in a "punchy," sales-y, formulaic way; also tied to parallelism).
- **Unnecessary mini-tables** that should just be prose.
- **Curly quotes / curly apostrophes** ("smart quotes"), especially inconsistent mixtures of curly + straight--with the caveat that Word/macOS/iOS and some tools also do this.
- **"Subject:" lines** copied into talk pages/messages as if they were emails or form submissions.
## **"Chatbot voice" tells (communication aimed at a user, not an encyclopedia)**
- **Customer-service collaboration language**: "Of course!", "Certainly!", "I hope this helps," "Would you like...," "let me know," "here is a...," plus a tone of being an assistant rather than an editor.
- **Knowledge-cutoff disclaimers + speculation about missing sources**: "As of [date]...," "up to my last training update...," "details are limited/scarce," "not widely documented," "based on available sources...," plus then "likely" speculation (sometimes about personal life: "keeps a low profile," etc.).
- **Mad-libs templates / placeholder text left in**: bracketed blanks ("[Entertainer's Name]"), "INSERT_URL," "PASTE_*_HERE," and placeholder dates like "2025-XX-XX."
## **Wikipedia-specific markup tells (copy/paste artifacts)**
These are especially relevant when someone pasted LLM output into Wikipedia without cleaning it:
- **Markdown instead of wikitext** (e.g., ## headings, Markdown links, Markdown lists).
- **Broken / garbled wikitext** because the model isn't reliably fluent in templates and MediaWiki syntax.
- **Residual tool-marker artifacts** like or image bundle markers (OpenAI tool/pipeline remnants).
- **Reference/citation markup bugs** such as contentReference[oaicite:...], oai_citation, Example+1, plus interface tags like [attached_file:1] or <grok-card ...>.
- **Weird "attribution / attributableIndex" strings** that don't belong in Wikipedia prose.
- **Hallucinated or out-of-place categories** (redlink categories, wrong naming conventions like missing hyphens).
- **Hallucinated templates and template parameters** (especially plausible-sounding infoboxes that don't exist).
## **Citation-pattern tells**
- **Multiple broken external links** (e.g., several 404s in a brand-new draft), suggesting the links never existed.
- **Invalid ISBNs/DOIs** (bad ISBN checksums; unresolvable DOIs), pointing to hallucinated references.
- **Outdated access-dates** that are oddly old relative to when the edit/draft was created--especially if many citations share the same old date.
- **DOIs that resolve--but to unrelated articles** (the DOI "looks real," but it's for a different paper).
- **Book citations missing page numbers** (or pages that don't actually support the claim), especially for broad, "sounds right" references.
- **Incorrect/unconventional reference mechanics**: broken re-use syntax; odd footnote symbols like ↩; citations pasted in ways that don't match Wikipedia conventions.
- **Tracking parameters that give away AI involvement**: URLs containing utm_source=chatgpt.com, utm_source=openai, utm_source=copilot.com, or referrer=grok.com (note: this proves an AI tool touched the sourcing, not necessarily wrote the prose).
- **Named refs declared but never used** (list-defined references inside a <references> block causing "Cite error...not used in the content").
## **Miscellaneous workflow tells (Wikipedia editing behavior)**
- **Sudden shift in writing style / English variety** (e.g., a user's usual writing vs suddenly polished, or a mismatch between expected regional English and the prose's variety--often default American English).
- **Overly verbose edit summaries** written like formal paragraphs that itemize policy compliance.
- **"Submission statements" embedded in Articles-for-Creation drafts** addressed to reviewers ("Reviewer note: this draft meets WP:... because...").
- **Pre-placed maintenance templates** that don't make sense (e.g., an AfC template already set to "declined," or random protection templates already present).
## **Historical (older-model) tells**
These were common earlier and are less reliable now, but still show up in older pasted text:
- **Didactic disclaimers** ("It's important/crucial to note...," "worth noting...," "may vary...") used as pseudo-neutral "editorializing."
- **Section/paragraph summaries** ("In conclusion," "In summary," "Overall...") and "Conclusion" sections in encyclopedic drafts.
- **Prompt-refusal boilerplate** ("As an AI language model...," apologies, "I can't do X, but I can...").
- **Abrupt cutoffs** (text ends mid-thought, sometimes from generation limits or bad copy/paste).
## **Extra rhetorical tells emphasized by the Dead Language Society piece**
These overlap with Wikipedia's "negative parallelisms / rule of three / em-dashes," but that essay frames them as **overused rhetorical devices** that become conspicuous when repeated:
- **Compulsive parallelism**: many sentences shaped the same way, with repeated mirrored structures.
- **Antithesis as a default move**: "not X; it's Y" and similar balancing moves used again and again to sound incisive.
- **Tricolon ("rule of three") everywhere**: triplets used to simulate cadence and completeness.
- **Signature punctuation choices** (especially em-dashes) used as a stylistic crutch.