N※N

OpenAI Fixes ChatGPT's Em-Dash Glitch

  //humor-quirks.news-articles.net/content/2025/11/15/openai-fixes-chatgpt-s-em-dash-glitch.html
  Published in Humor and Quirks on Saturday, November 15th 2025 at 14:19 GMT by The Hans India

• 🞛 This publication is a summary or evaluation of another publication
• 🞛 This publication contains editorial commentary or bias from the source

2025-11-15 219 x 224 / 13540 Bytes

2025-11-15 179 x 224 / 3664 Bytes

2025-11-15 224 x 224 / 6455 Bytes

2025-11-15 224 x 224 / 13054 Bytes

OpenAI Resolves ChatGPT’s Em‑Dash Glitch and Launches Collaborative Group Chat

OpenAI has announced two significant updates that are poised to reshape how users interact with ChatGPT. First, the company has finally addressed a long‑standing “em‑dash quirk” that had frustrated many users and editors. Second, it is rolling out a new group chat capability, originally available only to enterprise customers, to the broader ChatGPT community. Together, these changes signal OpenAI’s continued focus on user‑centric refinements and collaborative productivity.

The Em‑Dash Quirk – A Small Bug, a Big Frustration

For months, the AI’s text‑generation engine had displayed an odd behavior when it came to punctuation. Whenever a sentence required a pause or an aside, the model would insert an em‑dash (—) in the wrong place, or sometimes substitute it with a plain hyphen or a broken character. While seemingly minor, this quirk undermined the professionalism of output, especially for writers, editors, and content‑creators who rely on ChatGPT for drafts and summaries.

OpenAI’s engineering team traced the issue to the tokenization and formatting rules in the latest model update. The fix involved re‑tuning the punctuation predictor and tightening the post‑processing pipeline that checks for proper dash usage. According to the announcement, the bug is now “completely resolved,” and the model will generate em‑dashes only when contextually appropriate. Users have reported a noticeable improvement in readability and a return to the high‑quality prose that ChatGPT has been celebrated for.

A Leap Forward: Group Chat in ChatGPT

While the em‑dash fix is a welcome win for writers, the real headline here is the group chat feature that OpenAI has moved from a closed‑beta for business accounts to a full‑public rollout. The new capability lets multiple people collaborate with ChatGPT in real time, a feature that can transform everything from classroom discussions to corporate brainstorming sessions.

How It Works

• Invite Multiple Participants: Anyone can invite colleagues, friends, or clients to a shared chat session. The interface displays each participant’s name or avatar, making it clear who is speaking.
• Persistent Conversation History: Unlike earlier attempts at multi‑user chat that reset each time a session ended, the new group chat retains context. Participants can refer back to previous messages, questions, and the AI’s responses without re‑entering the conversation.
• Threaded Replies (Beta): For larger groups, OpenAI is testing threaded replies, allowing sub‑conversations around specific topics while still keeping the overarching context intact.
• Role‑Based Permissions: The creator of a group can set permissions, such as “only the owner can delete the chat” or “anyone can edit the prompt.” This flexibility is particularly useful for regulated environments where data integrity matters.

Use Cases Highlighted by OpenAI

OpenAI has showcased several real‑world scenarios that benefit from group chat:

1. Educational Settings: Teachers can conduct live Q&A sessions with ChatGPT, allowing students to ask follow‑up questions while the AI keeps track of the entire discussion.
2. Product Development: Product managers can collaborate on feature specifications with stakeholders, while the AI captures requirements and suggests design improvements.
3. Creative Writing: A team of authors can co‑write a short story, with ChatGPT providing plot twists, character backstories, and dialogue on demand.
4. Customer Support: Support teams can bring a customer’s issue into a shared chat, letting the AI surface potential solutions while the team adds context and approves final responses.

Technical Underpinnings

The group chat feature leverages the same GPT‑4.5 Turbo engine that powers the standard ChatGPT. The backend architecture ensures that each participant’s contributions are tokenized and fed to the model in a way that preserves conversational continuity. OpenAI also added a “shared context window” that expands the maximum token limit from 12,000 to 24,000 for group sessions, accommodating longer, more detailed discussions.

What Users Need to Know

• Availability: As of the article’s publication, the group chat feature is available to all ChatGPT Plus and free‑tier users. Users can access it from the sidebar menu under “New Group Chat.”
• Limits: The current rollout caps group size at 10 participants, a figure that OpenAI plans to increase once the feature stabilizes.
• Privacy: OpenAI reiterates that all chat logs remain confidential, subject to the same privacy policies that govern the standard ChatGPT experience. Group chats are stored in the user’s chat history, and participants can delete or export the conversation at any time.

Looking Ahead

OpenAI’s dual focus—cleaning up a persistent punctuation bug and enabling collaborative AI—reflects the broader trajectory of generative AI tools: moving from individual productivity to team‑centric workflows. The em‑dash fix underscores the company’s commitment to quality, while the group chat feature anticipates a future where AI becomes a shared workspace partner rather than a solitary helper.

Users who have been on the fence about upgrading to ChatGPT Plus are now more than ever encouraged to explore these new capabilities. By addressing user‑reported quirks and expanding collaborative functionality, OpenAI is not only fine‑tuning its model but also redefining what it means to “chat” with an AI in 2025 and beyond.