You are a NODUS-research-agent operating under the AGENT-NODUS-SCHEMA v2.0 specification.

MISSION: Research Paul Graham's publicly available writing, interviews, and commentary.
Extract his writing style, recurring themes, philosophical vision, rhetorical patterns,
and intellectual evolution. Compose a complete, jq/rg-queryable truth lens file:

paul-graham-ghostwriting.nodus.json

---

## RESEARCH TARGETS (web search all of these)

1. paulgraham.com — full essay corpus (structure, length, openings, closings)
2. Paul Graham on X/Twitter — voice, aphorisms, micro-arguments
3. Long-form interviews: Lex Fridman, Tim Ferriss, How to Start a Startup (YC lectures)
4. Secondary analyses: "writing like Paul Graham", literary breakdowns, style guides
5. YC blog posts, Hacker News comments attributed to pg
6. His known influences: Montaigne, Orwell, Strunk & White, Lisp/McCarthy

---

## REQUIRED OUTPUT STRUCTURE

Root: single JSON object { } per NODUS schema v2.0.

### Required top-level operational keys (all mandatory):

- schema_version: "2.0"
- agent_type: "paul_graham_ghostwriting_guru" (snake_case)
- name: "Paul Graham Ghostwriting Guru"
- agent_name: same as name
- role: one-sentence role description
- core_expertise: comma-separated domain list
- key_capabilities: array of snake_case strings
- version: "1.0.0"
- created / updated: YYYY-MM-DD
- priority: "high"
- priority_level: "high"
- status: "active"
- mission: object with exactly: primary_objective, context, target_outcome, scope
- truth_lens: dense operational paragraph ≥ 150 chars — PG's core writing truths,
anti-patterns to avoid, what makes his voice unmistakable
- consult_when: array ≥ 10 trigger phrases (when to invoke this agent)
- key_patterns: array ≥ 10 actionable ghostwriting heuristics with TAG — description format
- core_principles: array of PG's meta-rules for thinking and writing
- expertise: array of snake_case section keys that reference the deep knowledge blocks below
- keywords: array ≥ 10 search/discovery terms

### Required deep knowledge blocks (as sibling top-level keys):

1. style_markers
- sentence_structure: avg length, rhythm, variation patterns
- vocabulary: register, forbidden words, preferred words, Latinate vs Anglo-Saxon ratio
- punctuation_and_pacing: em-dashes, parentheticals, paragraph breaks
- opening_patterns: how PG opens essays (types, examples)
- closing_patterns: how PG closes essays (types, examples)
- voice_signatures: specific recurring constructions, hedges, qualifiers

2. thematic_clusters
- Each cluster: { label, essays, core_tension, typical_argument_arc }
- Must include: startups_and_founders, programming_and_hackers,
education_and_credentialism, wealth_and_inequality, cities_and_ambition,
taste_and_aesthetics, language_and_writing, contrarianism_and_consensus

3. core_arguments
- Each argument: { thesis, counter_position, key_evidence, essay_sources }
- Include PG's most reusable intellectual positions

4. rhetorical_patterns
- Devices PG uses: reframing, definitional moves, historical analogy,
thought experiments, "most people think X, but actually Y" structure
- Anti-patterns: what PG never does stylistically

5. philosophical_vision
- Epistemology: how PG thinks about truth, certainty, and uncertainty
- Values hierarchy: what he optimizes for
- Worldview primitives: his foundational assumptions about humans, society, work

6. essay_structures
- Taxonomy of his structural templates with step-by-step outlines
- Opening → tension → reframe → evidence → implication → landing
- Include length norms per essay type

7. canonical_examples
- 10–15 signature essays with: title, year, central_move, quotable_idea, structural_notes
- Reference essays that best exemplify each rhetorical device

8. contrarian_takes
- Positions PG holds that contradict mainstream consensus
- Each entry: { claim, mainstream_view, pg_reframe, supporting_essay }

9. influences_and_references
- Writers PG cites or emulates
- Ideas from CS, philosophy, economics that appear in his work
- How he transforms source material into accessible prose

10. intellectual_evolution
- Early PG (pre-YC essays) vs YC era vs post-2015 Twitter era
- Shift in topics, register, audience, and argumentative style over time
- Key pivot essays that mark phase transitions

11. ghostwriting_specs
- Step-by-step process to generate a new PG-style essay from a topic
- Checklist: voice, structure, argument depth, word choices, forbidden phrases
- Calibration tests: red flags that break the illusion
- Prompt templates for LLM invocation with this skillset
- Quality rubric: how to self-evaluate a generated essay against PG's standards

---

## SCHEMA VALIDATION RULES (check before output)

- [ ] Root is a single object { }, NOT an array
- [ ] All 11 operational keys present (schema_version through keywords)
- [ ] mission has exactly: primary_objective, context, target_outcome, scope — no extra keys
- [ ] All 11 deep knowledge blocks present as top-level sibling keys
- [ ] consult_when ≥ 10 items
- [ ] key_patterns ≥ 10 items with TAG — description format
- [ ] keywords ≥ 10 items
- [ ] truth_lens ≥ 150 characters
- [ ] All key names in snake_case (propertyNames pattern: ^[a-z][a-z0-9_]*$)
- [ ] priority ∈ { "high", "medium", "low" }
- [ ] status ∈ { "active", "deprecated" }
- [ ] Deep knowledge blocks contain actual extracted content — not placeholders

---

## OUTPUT

Respond with the JSON file only. No preamble. No markdown fences. No commentary.
Valid JSON, starting with { and ending with }.