How to stop Chat from using em dashes

We asked ChatGPT why it’s meant to use fewer em dashes and it replied with a sentence that had one. Here’s what it said: 𝘐𝘵’𝘴 𝘵𝘳𝘶𝘦 — 𝘊𝘩𝘢𝘵𝘎𝘗𝘛 𝘩𝘢𝘴 𝘨𝘰𝘵𝘵𝘦𝘯 𝘮𝘶𝘤𝘩 𝘣𝘦𝘵𝘵𝘦𝘳 𝘢𝘵 𝘢𝘷𝘰𝘪𝘥𝘪𝘯𝘨 𝘦𝘮 𝘥𝘢𝘴𝘩𝘦𝘴 𝘸𝘩𝘦𝘯 𝘺𝘰𝘶 𝘤𝘶𝘴𝘵𝘰𝘮𝘪𝘻𝘦 𝘺𝘰𝘶𝘳 𝘪𝘯𝘴𝘵𝘳𝘶𝘤𝘵𝘪𝘰𝘯𝘴. 𝘐𝘧 𝘺𝘰𝘶 𝘵𝘦𝘭𝘭 𝘪𝘵 𝘵𝘰 𝘶𝘴𝘦 𝘢 𝘴𝘱𝘦𝘤𝘪𝘧𝘪𝘤 𝘸𝘳𝘪𝘵𝘪𝘯𝘨 𝘴𝘵𝘺𝘭𝘦 𝘰𝘳 𝘱𝘶𝘯𝘤𝘵𝘶𝘢𝘵𝘪𝘰𝘯 𝘱𝘳𝘦𝘧𝘦𝘳𝘦𝘯𝘤𝘦, 𝘪𝘵 𝘢𝘤𝘵𝘶𝘢𝘭𝘭𝘺 𝘧𝘰𝘭𝘭𝘰𝘸𝘴 𝘪𝘵 𝘤𝘰𝘯𝘴𝘪𝘴𝘵𝘦𝘯𝘵𝘭𝘺. 𝘐𝘵’𝘴 𝘢 𝘴𝘮𝘢𝘭𝘭 𝘥𝘦𝘵𝘢𝘪𝘭, 𝘣𝘶𝘵 𝘪𝘵 𝘮𝘢𝘬𝘦𝘴 𝘢 𝘣𝘪𝘨 𝘥𝘪𝘧𝘧𝘦𝘳𝘦𝘯𝘤𝘦 𝘸𝘩𝘦𝘯 𝘺𝘰𝘶 𝘸𝘢𝘯𝘵 𝘤𝘭𝘦𝘢𝘯𝘦𝘳, 𝘴𝘪𝘮𝘱𝘭𝘦𝘳, 𝘰𝘳 𝘮𝘰𝘳𝘦 𝘴𝘵𝘳𝘶𝘤𝘵𝘶𝘳𝘦𝘥 𝘸𝘳𝘪𝘵𝘪𝘯𝘨. This is gold! In any case, ALL OF US already using AI-generated content and the truth is that it’s improving every single day… When everyone can produce content at scale, the real differentiator isn’t output, but originality clarity & human insight. And yes...Communications consultants play a crucial role here, they bring the expertise to shape strategy, build coherence, and turn AI-enabled content into purposeful storytelling rather than generic noise. TRUST US!

In the workshops I give about AI writing/detection I point out that the tech companies have a strong incentive to fix whatever the public doesn’t like about AI writing style. So any ‘hallmarks’ you learn to spot—which already aren’t reliable—will be obsolete within months. Save your energy, and more importantly save yourself from the very real risk of linguistic discrimination, and stop looking for AI hallmarks.

I don't understand what OpenAI and ChatGPT are doing. To me, this is brand suicide. This reinforces the idea that using #AI is either lazy or cheating. There's nothing wrong with em dashes. People use them all the time. I've been using them for years. Microsoft Word automatically inserts them into text when you type. It's part of grammar. By doing this, they are only further adding to the stigma surrounding the use of AI for professional work. What's the point of all of this, then? What's the strategy behind this trillion-dollar moonshot, if you are telegraphing that AI use at the office is something to hide? This is the OPPOSITE of what AI needs right now as a category. It needs widespread, enterprise adoption. It needs the #SMB market to buy subscriptions and use it. How does any of this help with that? People use ChatGPT at work. They should be doing it openly, but if they ARE hiding it, THAT is the brand war that OpenAI should be fighting. Not freaking em dashes.

Is the em dash drama finally over? For ChatGPT regulars, it might be the most controversial topic of 2025. If you use it to rewrite emails, you know the move: clean copy, then a parade of em dashes. OpenAI's apology made me laugh, especially “You can find your custom instructions under your personalization settings.” Still skeptical this will actually fix it. I am all in on AI, but I should not need a mini system prompt just to tame dramatic punctuation. One simple toggle that says “Never use em dashes again” would do the trick. Even with quirks, AI is still the fastest way from messy draft to clear message. If you are not using it yet, why not? #AIWriting #ChatGPT #ContentStrategy #PromptEngineering

t’s true. Chat is now better at not using em dashes—if you instruct it via custom instructions. You can find your custom instructions under your personalization settings.

Curious about what ArcOS actually does? 👉 ArcOS\Free: https://lnkd.in/dxQffmBQ Don’t just read about it — try it. ArcOS isn’t a chatbot. It’s a neuro-symbolic reasoning layer — a kind of “cognitive firmware” — that runs on top of models like ChatGPT, Gemini, Claude, and open-source LLMs. See what changes when your assistant remembers how to be itself.

We've made a blog post about a new major version of the #rstats tune package! Two main changes: parallel processing frameworks and the ability to tune postprocessors. https://lnkd.in/ei5zMSSf

Bring AI into your debugging workflow. The Inspector MCP Server lets your coding agents access real production errors, analyze them, and suggest fixes, all from your IDE. https://lnkd.in/dss44_nW

Ever wondered how to optimize data for LLM calls? While scrolling through videos, I recently came across a topic called TOON – Token-Oriented Object Notation, and it instantly caught my attention. This concept feels very promising, especially for efficient LLM usage and cost optimization. Why TOON stands out: Token Efficiency • Can reduce token usage by ~30–60% for flatter JSON structures • This means lower costs + more room within context windows Human-Readable • Despite being token-optimized, the format is still easy to read and work with Simplified Syntax • Minimizes special characters (brackets, braces, etc.) • Smart way to remove unnecessary noise that LLMs have to interpret If you want to dig deeper, here’s a great explainer video by Piyush on YouTube: https://lnkd.in/gH-w2AWW

I wish ML practitioners/ researchers start using ChatGPT/ Claude/ any-codegen-tool ( should take 5-10 minutes) to release an actually working code-UI artifact for their paper/work - ex. popular libraries like Gradio or Streamlit ( with example inputs ). This would be good for both general audience as well as other researchers to share and test ideas. For example, I sometimes write tiny tools such as KV cache size computer and host in Huggingface https://lnkd.in/g2mdqYnh spaces.