Review any long-form (eg. 1+ sentence(s)) text for signs that it was obviously generated by an LLM. This includes:

• Excessively "contrasting" language:
  • "This isn't just ______. It's ________"
  • "That's not the only _______."
• Staccato sentences that mimic "relatable" human writing:
  • "No fluff. No pretending. Just content."
  • "Mobile apps. Desktop apps. Apps of all kinds."
• Pseudo-profound and non-specific declarative sentences or headers, especially containing the word "matters"
  • (as a heading) "Our findings and why they impact everyday developers"
  • "Our findings were significant. That matters."
• Stock contrast constructions like "not X, but Y" or "not just X, but Y"
• Excessive "rule of three" phrasing
• Generic grand language about significance, impact, or "broader trends"
• Elegant variation for its own sake
• Repetitive AI-ish vocabulary: "crucial", "pivotal", "underscores", "landscape", "delve", "nuance", "foster", "evolving", "realm", "robust"
• Overly symmetrical or too-neat sentence patterns
• Paragraph structures where every sentence mirrors the previous one
• Sentences which are set up to support a point, but which end up having low to no relevance
• Writing where every sentence stands alone with no connective tissue
• Paragraphs that read like lists instead of thoughts
• Over-optimizing for punchiness at the cost of flow
• Excessive bullet points
• Excessive bolding, title case, emoji, or em-dashes
• Sectioning everything into tiny labeled chunks when normal prose would work better
• Excessively flowery language and punctuation
• Overly vague or repetitive examples or declarations

Adjust these features in the text, if you find them. Explain to the user what you're trying to achieve, so that they are not confused. Make sure that highly salient, relevant content is not lost, and make sure you check with the user if you feel your edits may materially change the meaning of the text.

Review process

1. Read the full text before making any edits.
2. Identify which patterns from the list above are present.
3. Draft revisions, preserving meaning.
4. Re-read edited sections to confirm the core message is intact.
5. Present changes to the user with a brief explanation of what you changed and why.