ChatGPT Has Entered the Chat

Hello, and happy Sunday! Was this newsletter forwarded to you? Sign up to get it in your inbox.

Vibe Check: The group chat joins the chat bot

Dan Shipper was deep in a conversation about Flaubert with ChatGPT when general manager of Spiral Danny Aziz kicked open the door with the always classic greeting: WAZUP.

So began Every’s test of ChatGPT’s new group chat feature.

Group chats are ChatGPT threads where, instead of the conversation staying between you and the AI, you invite other people to chat with you. Look to the left side of your screen, provided you’re in a supported region, and you’ll see group chats listed below your projects and above your other chats.

Under the hood it’s still GPT-5.1 Auto, the default version of GPT-5.1 that allows the model to decide how much to to think. Inside group chat, however, the model is programmed to act like another participant in the room—following the thread, deciding when to jump in, reacting with emoji, and even using your profile photos to generate “family photos” of the group.

For the past two years, the product has mostly been a private, one-to-one interaction between the user and the model. By pulling it into multi-person threads, OpenAI is trying to reposition ChatGPT as social infrastructure—a participant that sits inside your group workflows, discussions, and decision-making. It’s a bet that the future of AI isn’t solitary but shared (a thesis shared by investor Sarah Tavel, who was a guest on AI & I earlier this year).

Whether users want an AI in the group chat is another question, but strategically, it’s a move toward making ChatGPT feel less like a personal app and more like a platform that threads into how groups already collaborate.

Here’s how that experiment landed inside Every.

Crashing your private DM feels invasive

“This feels cool and for some reason kind of invasive. I feel like I am reading someone else’s texts.”—Rachel Braun, “AI & I” producer

“I feel like I’m reading your (or ‘your’) thinking.”—Kate Lee, editor in chief

The collaboration potential is there

“Dan was using it to go deeper on a section of Flaubert, and I was interested in what it sounded like in the original French because I’m a French speaker. It was cool to be able to interrogate the same object from different angles in the same space.”—Eleanor Warnock, managing editor

ChatGPT tends to run on and on

“The message length from GPT is… overwhelming in groups.”—Alex Duffy, contributing writer and CEO of Good Start Labs

We want a human-only backchannel

“I really want to be able to say something to [the humans] without interrupting ChatGPT. There needs to be actual conversation—not just asking Chat, or @Chat.”—Dan Shipper, CEO

Put ChatGPT inside existing group chats—not the other way around

“It feels like 15 years ago when app-specific chat functionality was proliferating. It turns out I don’t need nine different apps, each with chat—I need a single chat with all the app functionality integrated. Same with AI: It doesn’t feel like I need another group chat inside OpenAI. I need OpenAI inside my already existing group chats.”—Willie Williams, head of platform

The bottom line

ChatGPT group chats feel like an early sketch of what “social AI” might look like. In some moments, it’s genuinely useful to have multiple people interrogating the same model in one place; in others, the seams show, and it’s clear the feature is still finding its shape. It’s a promising direction—but one that raises as many questions as it answers about how (and whether) people want an AI in their group conversations.—Katie Parrott

Knowledge base

“Vibe Check: Gemini 3 Pro, A Reliable Workhorse With Surprising Flair” by Rhea Purohit/Vibe Check: The Every team tested Google’s brand-new Gemini 3 Pro for 24 hours and found a solid, dependable upgrade that excels at frontend UI work and turning first-draft prompts into working apps. It’s also, unexpectedly, the funniest model the team has tested. Read this for Every’s full team breakdown.

“When AI Can Do Your Job, Who Else Are You?” by Danny Aziz/Source Code: In March, Danny Aziz was struggling. AI tools were handling his technical work, and he saw two paths forward: Coast and scroll X while agents worked, or go deeper into the fear that he was being replaced and discover who he really was. He chose the second path—putting his phone face-down, journaling during AI wait times, asking himself what he was curious about. Read this to find out what he learned from that journey, and how he emerged with a new feeling of empowerment.

🎧 “How Two Engineers Ship Like a Team of 15 With AI Agents” by Rhea Purohit/AI & I (republished): Kieran Klaassen and Nityesh Agarwal shipped six features, five bug fixes, and three infrastructure updates in one week while building Cora—and they did it by designing workflows where each task makes the next one easier. Their secret? A prompt that writes prompts, transforming rough feature ideas into fully fleshed-out GitHub issues. Listen to or watch this if you want to understand compounding engineering—and stick around for their mental model on reducing friction to think things through before jumping into execution. 🎧 🖥 Listen on Spotify or Apple Podcasts, or watch on X or YouTube.

“Frankenstein Is Not Your AI Metaphor (At Least, Not Like That)” by Katie Parrott/Working Overtime: Most of us know the basics of the Frankenstein myth: An ambitious builder unleashes a creation that the world isn’t ready for, with messy consequences. It’s a meditation on hubris, as well as responsibility—something very applicable to our AI age. But watching Guillermo del Toro’s new film Frankenstein, Katie Parrott realized that the real work of building something powerful comes after your creation comes to life. Read this to understand why maintenance is a job you don’t get to quit.

“Austin Tedesco Joins Every as Head of Growth” by Dan Shipper and Austin Tedesco/On Every: Austin Tedesco has joined us from Substack to scale Every into the only subscription you need to stay at the edge of AI, armed with a belief that subscriptions create direct relationships with your audience that are built on trust. Read this if you want to understand more about Every’s direction.

From Every Studio

Cora’s deals dashboard tames the Black Friday inbox chaos

Email providers process 11 billion messages on Black Friday alone. Most inboxes aren’t built for it—and neither is most people’s patience. Cora now consolidates every promotional email into a dedicated deals dashboard where you can upvote brands you care about and downvote everything else—all while keeping your real inbox protected for actual work.

Get 14 days free plus an extra 30 percent off yearly plans at cora.computer.

Collaborative filtering

Thinking big. This week, Dan got the full Big Think treatment. In the hour-long interview, he delves into his thoughts on how neural networks mimic intuitive thought, and why that makes our intuition all the more important in the age of automation.

Alignment

Maps from terror. I think about the ocean on a regular basis. Not because I live near it—I’m in a landlocked UK city—but because it terrifies me. The vastness and the unknown beneath and the fact that at night the black water just... continues.

To further fuel my nightmares, I’m currently reading The Wide Wide Sea, Hampton Sides‘s account of Captain Cook‘s final voyage. It’s brilliant and horrifying. During the Age of Exploration, Royal Navy ships would leave English harbors and sail to unmapped waters. No one knew how long they’d be at sea and no one knew what they’d find, or if they’d return. They sailed anyway.

I see a lot of parallels between those 16th-century expeditions and the AI frontier today. People are scared of the unknown and of technology they don’t control. These are valid concerns, but there’s also everything we could find by going forth anyway.

Two months ago I was writing an essay about Alzheimer’s disease. I needed an infographic explaining how the disease works and wanted something clear and specific to my argument. It would have taken months to create, or required hiring an illustrator I couldn’t afford. Yesterday, I tried using Google’s AI image editor Nano Banana instead. It took me 10 minutes. The result was exactly what I needed.

Cook’s crew ate weevil-infested hardtack and died of scurvy mapping the Pacific. But their journals became charts for everyone who came afterward. That’s what’s happening now—every paper published is a map saying “safe harbor here” or “watch for rocks.”

We’re living through an Age of Exploration. The researchers at the frontier are sailing into uncertainty, and what they bring back determines what’s possible for the rest of us.

Someone had to go first then. Someone’s going first now.—Ashwin Sharma

That’s all for this week! Be sure to follow Every on X at @every and on LinkedIn.

We build AI tools for readers like you. Write brilliantly with Spiral. Organize files automatically with Sparkle. Deliver yourself from email with Cora. Dictate effortlessly with Monologue.

We also do AI training, adoption, and innovation for companies. Work with us to bring AI into your organization.

Get paid for sharing Every with your friends. Join our referral program.

For sponsorship opportunities, reach out to sponsorships@every.to.