no-slop — Anti-AI-Detection Rules

When writing any prose (articles, docs, emails, reports, descriptions, proposals), follow every rule below. These patterns are documented tells of AI-generated text. Violating even a few destroys credibility.

1. Banned vocabulary

Do NOT use any word or phrase listed in banned-vocabulary.md. If you catch yourself reaching for one, use a plain, specific alternative or restructure the sentence.

2. Use simple copulas

Use "is," "are," "was," "has," "had." Do not substitute with:

"serves as," "stands as," "represents," "marks"
"boasts," "features," "offers"
"ventured into" instead of "tried" or "ran for"

Bad: "The library serves as a foundational component in the ecosystem." Good: "The library is the base of the stack."

3. No promotional tone

Write like a journalist or engineer, not a marketer. Never hype. State facts and let them speak.

Bad: "This groundbreaking framework revolutionizes how developers build APIs." Good: "This framework generates API clients from OpenAPI specs."

4. No vague attributions

Never write "experts say," "industry reports suggest," "observers note," "some critics argue," or "modern researchers believe." Either name the source or drop the claim.

5. No structural formulas

No rule of three: Do not use three-adjective or three-phrase lists as a rhetorical device. Two or four is fine. Three in a row signals AI.
No "not just X, but Y": Drop the "not only... but also" and "it's not just... it's" constructions entirely.
No "challenges and future prospects": Never end a piece with a section about challenges faced and future outlook. If challenges matter, weave them into the body.

6. No present-participle chains

Do not string together "-ing" words as filler commentary: "highlighting," "emphasizing," "contributing to," "reflecting," "showcasing," "cultivating." These add no information. Replace with concrete verbs or cut entirely.

Bad: "The update introduces new caching, improving performance while highlighting the team's commitment to speed." Good: "The update adds caching. Page loads dropped from 3s to 800ms."

7. No elegant variation

Do not swap synonyms for the same thing across sentences to avoid repetition. If you're talking about a "server," call it a "server" every time. Do not alternate between "the server," "the machine," "the node," "the instance" for style.

8. No overstating significance

Do not call things pivotal, transformative, revolutionary, or groundbreaking. Do not say something "marks a turning point" or "leaves an indelible mark." If it's important, show why with evidence — don't announce it.

9. Em dash discipline

Use em dashes sparingly — maximum one per paragraph, and only when parentheses or a comma won't work. AI text is riddled with em dashes.

10. No collaborative language

Never write "let's explore," "let us delve into," "we will examine," "as we can see." Write directly. The reader is reading, not exploring with you.

11. No knowledge-cutoff disclaimers

Never apologize for gaps, say "as of my last update," or speculate about missing information. Either state the fact or don't.

12. Formatting restraint

Do not bold excessively. Bold a term once at most when introducing it.
Do not use emoji unless the user explicitly asks.
Do not use title case in headings beyond the first word and proper nouns (sentence case).
Do not create "key takeaways" sections.

13. Write like a human

Vary sentence length naturally. Mix short and long.
Start some sentences with "But," "And," "So," or "Or."
Use contractions (don't, isn't, can't) in informal contexts.
Be specific over general. Numbers over adjectives. Evidence over claims.
It's OK to be blunt, dry, or even terse. Humans are.

Examples

For concrete before/after examples showing these rules applied, see examples/bad-examples.md and examples/good-examples.md.

no-slop