Prose is free now

AI can generate fluent, grammatical, well-structured text on any topic. The marginal cost of generating 2,000 words of coherent AI-generated text is negligible. As a16z puts it: the marginal cost of creation is approaching zero.

This changes which parts of content creation are scarce and valuable:

EraScarceAbundantValue lives in
Pre-AIProseIdeas, facts, artifactsThe writing itself
Post-AIIdeas, facts, artifactsProseWhat the prose carries

Prose is the medium, and the value lies in specific elements the reader can extract and use. I call those elements substantives.


Substantives

A substantive is a discrete, reusable unit of content that carries standalone value for the reader. Examples include definitions, checklists, code samples, and diagrams that can be reused or referenced without the surrounding explanation.

The extraction test: Assume a reader skims your piece in 30 seconds and extracts 3-5 things to save or share. Identify what they would extract.

  • A checklist: extractable, substantive
  • A screenshot: extractable, substantive
  • A paragraph of explanation: usually not, tissue
  • A definition: extractable, substantive
  • A transition sentence: not extractable, tissue

Pieces built only from connective prose are harder to reuse or reference later.


Why This Matters Now

Readers quickly filter out low-value content. The backlash is real: "AI slop" entered the lexicon, and mentions increased ninefold in 2025. Platforms are responding: YouTube stopped paying for AI slop, and Pinterest added filters to hide AI content.

Readers ignore writing that shows these patterns:

  • Generic claims without specific evidence
  • Fluent prose without anchoring details
  • Structure without substance

Many AI-generated listicles, self-described comprehensive guides that say nothing, and posts with interchangeable advice contain no specific, reusable elements.


The Taxonomy

Substantives can be grouped by what the reader gains from them:

CategoryReader getsExamples
UtilitySomething to useCode snippet, checklist, template, prompt, procedure
KnowledgeSomething to knowDefinition, taxonomy, framework, distinction
ProofReason to believeScreenshot, benchmark, citation, timestamp, specific name
PerspectiveNew way of seeingReframe, analogy, decomposition, contrast
ConnectionOrientation in spaceExternal link, comparison table, prerequisite pointer
ExperienceFelt understandingWorked example, failure story, interactive demo
ShortcutCompressed wisdomHeuristic, pattern, anti-pattern, threshold, trade-off
VocabularyWords to think withCoined term, precise definition, distinction pair

Not every piece needs all eight. Most pieces have 2-3 primary value types.


The Workflow

One common workflow:

  1. Define the main message
  2. Outline the structure
  3. Draft and refine the prose
  4. Add supporting evidence

Substantive-first:

  1. Define what the reader should have when they finish
  2. Identify substantives that provide that value
  3. Order them into a sequence
  4. Add prose that connects them

Identify and design the substantives first, then write the prose that presents and connects them.


Checklist Before Publishing

Before publishing, scroll through looking only at visual breaks and extracted elements:

  • Is there a substantive every 300-500 words?
  • Could someone extract 5+ useful things?
  • Is there at least one visual (screenshot, diagram, table)?
  • Is there at least one structural (checklist, framework, definition)?
  • Are there inline substantives (links, specific names, numbers)?

If the piece contains mostly transitions and explanation with few concrete elements, revise it by adding more substantives or removing nonessential prose.


How Substantives Differ from Arguments

Traditional essays rely on arguments: claims supported by reasoning that the reader is expected to follow. Substantive-first content is built around artifacts, like checklists or definitions, that a reader can use without reading the rest of the piece.

ArgumentArtifact
"X is true because Y"Screenshot showing X
"You should do X"Checklist for doing X
"X works better than Y"Benchmark comparing X and Y
"X means Y"Definition: X = Y

Arguments depend on the reader accepting your reasoning. Artifacts like checklists or definitions provide value directly.


The Reusable Artifact Test

To make a document reusable as a reference, ensure that it is:

  1. Extractable - Elements can be pulled out and used independently
  2. Stable - Content doesn't rely on context that will change
  3. Addressable - Specific sections can be pointed to
  4. Machine-readable - AI can parse and reference it

Treat the document as a reusable artifact when automated tools can reliably parse its structure, you can copy sections into other documents without rewriting, and you can link directly to specific sections.


The shift: start with what the reader should HAVE, not what you want to SAY.

Design the substantives first, then write the prose, and test the result by checking what can be extracted.


Reference: Exhaustive Substantive Types

Utility — Reader Can Use This

SubstantiveWhat Reader GetsWhen To Use
Code snippetExecutable capabilityAny technical content where reader might implement
TemplateFill-in-the-blank starting pointProcess or document the reader will create themselves
ChecklistVerification tool, nothing forgottenAny multi-step process with failure modes
ProcedureStep-by-step execution path"How to" content of any kind
Prompt textReady-to-use AI invocationAI-related content, workflow content
ConfigurationWorking setup they can copyTechnical content with setup requirements
Query/commandExecutable one-linerTechnical content, data content
Calculation/formulaComputable relationshipQuantitative decisions, estimation
Decision treeNavigation through choicesComplex decisions with branches

Knowledge — Reader Now Knows This

SubstantiveWhat Reader GetsWhen To Use
DefinitionPrecise meaning of termAny time you use a term that could be ambiguous
TaxonomyClassification systemDomains with multiple types, options, or categories
FrameworkMulti-part mental modelComplex domains requiring structured thinking
FactVerified true statementFoundational claims, surprising truths
DistinctionDifference between two thingsCommonly confused concepts
RelationshipHow X connects to YSystems, dependencies, causes
Counter-exampleCase where intuition failsCorrecting common misconceptions

Proof — Reader Is Convinced

SubstantiveWhat Reader GetsWhen To Use
ScreenshotVisual evidence something existsAny claim about a system, UI, output
BenchmarkMeasured performance dataClaims about speed, quality, comparison
Log outputMachine-generated evidenceClaims about system behavior
Repo linkInspectable source of truthAny claim about code you've written
CitationAuthority backingClaims reader might doubt, contested topics
Testimonial/quoteAnother person's attestationClaims about user experience, reception
ReplicationSomeone else got same resultScientific or technical claims
TimestampWhen something happenedNarrative claims, incident reports
Specific nameWho/what specificallyAny claim that could be vague

Perspective — Reader Sees Differently

SubstantiveWhat Reader GetsWhen To Use
Reframe"Not X, but Y" shiftCorrecting common framing errors
AnalogyUnderstanding via familiar domainComplex concepts, cross-domain transfer
Lens/modelWay of analyzing thingsRecurring situations reader will face
ContrastBefore/after, old/newChange, improvement, evolution
DecompositionBreaking whole into partsComplex wholes that seem monolithic
SynthesisCombining parts into wholeScattered elements that form pattern
Scale shiftZooming in or outWhen reader is stuck at wrong altitude

Connection — Reader Is Oriented

SubstantiveWhat Reader GetsWhen To Use
External linkPath to deeper/related contentAny mention of external work
Comparison tableRelative positioning of optionsDecisions between alternatives
Influence chainWhere ideas came fromNovel concepts, intellectual positioning
Contrast with alternativeHow this differs from XCompetitive positioning
Prerequisite pointerWhat to learn firstContent with dependencies
Next step pointerWhere to go after thisContent that's part of larger journey
Ecosystem mapHow this fits in larger landscapeTools, frameworks, communities

Experience — Reader Felt This

SubstantiveWhat Reader GetsWhen To Use
Interactive demoLearning by doingConcepts that must be experienced to understand
Worked exampleWatching the processSkills, procedures, problem-solving
Narrative incidentVicarious experienceLessons that come from living through something
Failure storyLearning from pain (yours)Warnings, cautionary knowledge
Dialogue excerptWitnessing an exchangeInterpersonal dynamics, debates
Sensory detailGrounding in physical realityAbstract content that needs anchoring
ExerciseActive practiceSkills that require repetition

Shortcut — Reader Has Compressed Wisdom

SubstantiveWhat Reader GetsWhen To Use
HeuristicRule of thumbRepeated decisions, common situations
PatternRecurring structure with nameSituations reader will encounter multiple times
Anti-patternWhat to avoid and whyCommon failure modes
ThresholdWhen X becomes YDecisions that depend on quantity
Priority orderWhat to do firstResource-constrained situations
Trade-offWhat you give up for whatDecisions with real costs
Warning signIndicator of problemDiagnostic situations

Vocabulary — Reader Can Think With This

SubstantiveWhat Reader GetsWhen To Use
Coined termNew word for previously unnamed thingWhen existing vocabulary fails
Precise definitionSharpened meaning of existing termWhen term is used sloppily
Distinction pairTwo words that mark important differenceWhen one word conflates two things
Acronym/abbreviationCompressed referenceConcepts that will be referenced repeatedly
CatchphraseMemorable crystallizationCore insight you want reader to retain and repeat