Level 0
signal
form + contrast
Create a perceivable difference.
Fails when: The reader cannot detect a meaningful distinction.
LVWGS · version 1.0.0
Writing, taken apart to first principles.
This is a decision system, not a style guide. It breaks writing into eleven levels — from a perceivable mark to a governed content system — and states what happens at each level as machine-readable logic: if this condition holds, do that; otherwise, do this instead.
Every rule is labelled with how hard it binds. Hard constraints must pass or the text is blocked. Context defaults apply unless you can name the exception. Heuristics are ranked, never forced.
What is in the corpus
289
Operational rule records
228
Base rules
56
LAKA transformation states
30
Source files
Counts derive from the corpus itself at load time, not from the manifest's declarations. Source: the 30 JSON files served at /v1/files, version 1.0.0, generated 2026-08-31.
The spine
Eleven levels, each with one job
signal → grapheme → morpheme → word → phrase → clause → sentence → paragraph → section → document → content system
Level 1
grapheme
mark + writing-system value
Represent a letter, digit, symbol, or punctuation mark.
Fails when: The mark is missing, confusable, or malformed.
Level 2
morpheme
form + minimal meaning_or_function
Carry the smallest grammatical or lexical contribution.
Fails when: Inflection or word formation expresses the wrong relationship.
Level 3
word
one_or_more_morphemes + lexical_role
Name, relate, modify, quantify, or operate.
Fails when: The selected word is inaccurate, ambiguous, inaccessible, or inconsistent.
Level 4
phrase
head + optional_dependents
Build a unit around a noun, verb, adjective, adverb, or preposition.
Fails when: The head-dependent relationship is unclear or misplaced.
Level 5
clause
predicate + participants + optional_circumstances
Represent an event, state, relation, or proposition.
Fails when: The actor, action, object, time, condition, or logical status is missing or confused.
Level 6
sentence
one_or_more_clauses + relation + closure
Deliver one manageable assertion, question, command, or exclamation.
Fails when: The sentence is incomplete, fused, overloaded, or logically misjoined.
Level 7
paragraph
controlling_move + development + coherence
Advance one main rhetorical move.
Fails when: The point is buried, unsupported, mixed with another point, or disconnected.
Level 8
section
reader_question + organized_answer + navigation
Resolve one meaningful reader need or document function.
Fails when: The section has no clear question, sequence, or usable label.
Level 9
document
purpose + audience + ordered_sections + evidence + action_path
Change what a reader knows, feels, decides, or does.
Fails when: The document is correct locally but fails its audience or purpose globally.
Level 10
content system
documents + taxonomy + reuse + governance + feedback
Produce consistent, maintainable communication across contexts.
Fails when: Rules conflict, drift accumulates, content duplicates, or feedback never changes the system.
How it decides
Three ways in
Look up a rule
Filter 228 rules by domain, layer or how hard they bind. Each one shows its condition tree, both branches, its exceptions, and the sources behind it.
Run the engine
Paste a draft. The engine analyses it, selects the rules that apply, evaluates each one, and separates what blocks from what is merely ranked.
Call the API
Every endpoint is public JSON. Read the corpus, resolve a rule set for a context, or evaluate text from your own tooling.
Read this before you use it
The system decides how hard a rule binds — not that everything should be plain.
This is a decision system, not a universal command to make all prose plain or uniform. Apply genre, audience, stakes, dialect, locale, and deliberate exceptions while preserving truth, safety, accessibility, and intended meaning.
Hard constraint
must pass or block
Context default
apply unless a documented exception improves the intended outcome
Heuristic
score and rank
Creative option
offer as a controlled alternative