Prompt Guide to Stop AI Models Sounding Mannered

X postAlex VeremeyenkoResource

Alex Veremeyenko shares a prompt attributed to Anthropic designed to eliminate overly florid and performative writing in Claude, Fable 5.1, GPT-5.6 Sol, and GPT-6 Astra.

Key points

  1. Critique of 'mannered prose': AI models frequently swap plain assertions for unnecessary metaphors and stylistic flourishes.
  2. Contrasted examples: 'A dial worth turning' instead of 'a parameter worth varying', or 'this point earns its keep' instead of 'this point still matters'.
  3. Flourishes create friction for the reader and drag in unintended connotations that reduce precision.
  4. The directive's core fix is straightforward: always use a literal phrase whenever one is available.
  5. The author suggests pasting this prompt into model custom instructions permanently.

Why it matters

LLM output often suffers from identifiable, formulaic metaphors that obscure technical meaning. Injecting explicit guidelines against performative language helps ensure generated copy remains concise, professional, and clear.

Analysis

Rather than relying on abstract negative constraints, this instruction explains the mechanics behind bad stylistic choices and provides direct before-and-after contrasts. This structure helps guide LLM attention toward functional, literal communication.

Original post

Alex Veremeyenko

@alex_verem · 9 September 2026

Anthropic shared this brilliant prompt that stops Claude from sounding like Claude. It improves writing with Fable 5.1 as well as GPT-5.6 Sol plus GPT-6 Astra. Paste it into your custom instructions and leave it there. Mannered prose substitutes metaphor and flourish for t.co/lLcRqMDTr0

419 likes16 replies852 bookmarks41.8k views

View on X
More posts in the thread (2)
  1. Source: t.co/mjBciLOBHC
  2. I curate the best AI projects actually solving real problems. Ecology. Medicine. Agriculture. Not chatbot tricks. Follow for the AI that matters 👇 t.co/YAht3k76LL
Selected replies (5)
  • @Tsuchida @alex_verem @grokしかし日本語だと不自然な表情ですね。どう言い換えればいいかな
  • @retarded1whale @alex_verem Kinda sounds just as terrible of AI robot jargon honestly , nobody says a dial worth turning or this point earns its keep , good prompt structure though I think gotta delete the examples and write our own based on how we actually speak ,
  • @sebuzdugan @alex_verem which anthropic source published this prompt, without a link the claim is unverifiable
  • @Terabhaitoxic @alex_verem @grok Wow such a genius to first introduce watermarks on claude and then give methods to remove that just so uses more tokens in background
  • @KuzkaaKuzkaa888 @alex_verem Say what you mean. That’s the whole prompt.“A dial worth turning” is the model performing. “A parameter worth varying” is the job.Metaphors sneak in meanings you didn’t pick. Literal sentences don’t.Paste it into custom instructions and leave it. One-shot won’t kill the habit.If

Mentioned links

Links and tools from this entry

Summary generated by AI (Gemini 3.8 Flash, 27 September 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • X post:Meng To

    no-ai-slop: Skill removes repetitive AI writing patterns

    Meng To shares a workflow to counter generic AI writing, highlighting Peter Yang's open-source repository 'no-ai-slop', which strips more than 20 repetitive LLM clichés and robotic phrases from text.

    1.8kbookmarks126.6kviews

    AI & LLMs· Tool

  • Article:Peter Yang

    Open-Source Skill to Strip AI Slop and the 25/50/25 Rule for Writing

    Peter Yang has open-sourced an AI skill called /no-ai-slop for harnesses like Claude Code and Codex. The tool detects and removes over 20 repetitive patterns commonly produced by generative models. Alongside the release, Yang outlines a 25/50/25 writing process to retain editorial control.

    AI & LLMs· Tool

  • Article:Anthropic

    Anthropic Prompting Best Practices for Current Claude Models

    Anthropic outlines prompt engineering best practices for its Claude model line-up, spanning Claude Haiku 4.5 through Claude Fable 5.1 and Claude Mythos 5.1. The guidance covers model-specific nuances, XML structuring, few-shot examples, system roles, and optimal placement for long contexts.

    AI & LLMs· Resource

  • X post:Boris Cherny

    Boris Cherny on Prompting: Treat Claude Like a Coworker

    Boris Cherny expresses surprise at people being astonished by his simple prompting style for Claude. He argues that most tasks do not require rigid scaffolding; users should simply set a goal and speak to the model as they would to a coworker.

    5.8kbookmarks603.1kviews

    AI & LLMs· Opinion

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.