Evaluation of https://minikin.me/blog/cargo-crap against three writing style guides: humanizer, stop-slop, and writing-clearly-and-concisely (Strunk's Elements of Style).
The post about cargo-crap displays several characteristic AI writing patterns across all three style guides. Quotes are pulled verbatim from the post.
Negative parallelisms (pattern #9) - heavy use:
- "The goal is not to make every number perfect or to ban change. The point is to make risky change visible..."
- "It does not replace review, but it gives review a sharper starting point."
- "Good tools do not replace thinking. They make thinking easier to focus."
- "Software quality is not only about whether the code compiles or whether tests exist. It is about whether we can safely change the system tomorrow."
Rule of three (pattern #10):
- "Memory safety. Thread safety. Ownership. Lifetimes. Exhaustive matching. Strong types." (forced staccato list)
- "Parsers, state machines, protocol handlers, validation logic, and compatibility layers"
- "another fallback, another special case, or another compatibility branch"
- "generate code, refactor functions, add tests, update APIs"
Em dash (pattern #13): "It is a signal --- a useful one."
Boldface overuse (pattern #14):
- "Change Risk Anti-Patterns"
- "boundary for AI-assisted development"
- "AI agents"
- "The real problem is untested complexity"
Copula avoidance (pattern #8): "cargo-crap serves as a boundary..."
Quotable pull-quote endings: Two blockquoted aphorisms ("Move fast, but measure the risk you are adding." / "We may have existing problems, but new changes should not make them worse.") plus paragraph closers like "every change becomes a small act of faith."
False agency (inanimate subjects doing human verbs):
- "Rust makes many bugs impossible"
- "Coverage shows..." / "Complexity shows..."
- "CRAP combines both signals to highlight code..."
- "an agent adds another fallback... to keep the current behavior working"
- "cargo-crap makes silent increases... visible"
Binary "not X, but Y" structures: "The real problem is untested complexity, not complexity itself." Plus the negative parallelisms listed above.
Adverb spray: perfectly, completely, naturally, gradually, automatically, happily, intentionally, quietly, simply.
Punchy one-liner paragraph endings: Multiple paragraphs land on quotable one-liners ("that is where code changes become dangerous"; "It is a signal --- a useful one"; "Move fast, but measure the risk you are adding"). Metronomic.
Throat-clearing meta-joiners: "Here is the kind of shape cargo-crap is meant to make visible:" / "That is the practical value."
Rule 11 - Put statements in positive form: The post leans on negative framing constantly ("It is not a replacement...", "It does not understand...", "It does not prove...", "does not pretend that...", "is not meant to shame...").
Rule 13 - Omit needless words: "You do not need to memorize the formula." / "That speed cuts both ways." / "That last category is what matters because that is where..." Padding sentences that restate the prior point.
Rule 10 - Active voice: Several passive or pseudo-passive constructions ("are becoming part of", "can be misleading", "is meant to make visible").
The post has a real first-person author ("I built it as one small guardrail..."), concrete artifacts (an actual formula, working CLI examples, a real PR link, a code sample with classify_event), and a specific technical claim. That grounding is what a pure LLM dump usually lacks.
This reads as human-edited LLM output or a human writer who has absorbed LLM cadence. The technical scaffolding (formula, code, CI snippet, PR link) is genuine, but the connective prose - the negative parallelisms, the forced threes, the bolded buzz-phrases, the aphoristic blockquotes, the "Final thoughts" wrap-up that restates the thesis - hits nearly every pattern the three guides flag. Score: heavily slop-flavored, not bottom-of-the-barrel slop.