ChatGPT finally fixes the em-dash

🔍 AI Idea of the Week: ChatGPT Finally Fixes the Em-Dash

OpenAI just rolled out a small but satisfying update: ChatGPT will finally stop overusing em dashes when you tell it not to. 

Sam Altman says the model now actually respects style preferences like “no em dashes”. 

It’s a tiny fix, but it makes AI-written text feel more natural and a lot less obviously AI.

⚡ AI Signals: The Updates You Shouldn’t Miss

Google’s just launched Gemini 3 Pro 

Tests show big improvements in multimodal understanding, reasoning, and proactive follow-up suggestions - a solid jump forward for Google’s flagship model. 

You’ll notice it runs faster and keeps context more consistently, even across multi-step conversations.

ChatGPT gets group chats

OpenAI launched group chats for up to 20 people, so you can now brainstorm, plan, or collaborate with AI and humans all in one thread.

It’s rolling out to logged-in users in select regions (Japan, New Zealand, South Korea, Taiwan), and you can start one by tapping the people icon and sharing the link with your group.

ChatGPT Atlas gets useful upgrades

OpenAI’s AI-powered browser now has vertical tabs, iCloud passkeys, and the option to set Google as your default search, making it feel much more like a daily-driver browser.

TikTok adds an AI-content control slider

TikTok now lets you choose how much AI-generated content you want in your feed, plus new invisible watermarks to help label AI-made videos.

ChatGPT for Teachers is now available

OpenAI released a free educator-focused version of ChatGPT (free until June 2027), designed to help teachers with lesson planning, creating materials, and working with files and images.

🛠 AI Tool We’re Testing: NanoBanana Pro

Google just launched the Pro version of Nano Banana 🤩

You can combine up to 14 images, include several people in one scene, tweak lighting and camera angles, and even add clean, readable text directly onto the image (think posters, ads, or social graphics).

What’s upgraded:

✅ Image and character consistency across multiple scenes
✅ Clean, readable text rendered directly inside images
✅ Fine control over lighting, angles, lenses, and camera styles
✅ High-resolution output (up to 4K)
✅ Blending multiple reference images into one coherent result

How to try it:

1. Open Gemini in your browser or app and look for Create images.

2. Pick Nano Banana.

3. Write a clear prompt, and (optionally) upload reference photos you want it to match in style, layout, or people.

4. Refine the result by asking for tweaks: change lighting (day/night), adjust angle, move objects, update colors, or improve the text on the image.

Thanks for reading, hope you have a fantastic week!

Best,

Darius