OpenAI Makes ChatGPT “Warmer” After GPT-5 Disaster
4IR - Daily AI News
Welcome back to 4IR. Here’s today’s lineup:
OpenAI makes ChatGPT “warmer” after GPT-5 disaster - GPT-5.1 drops three months after users revolted against the original GPT-5, promising better vibes, faster thinking, and personality controls after people literally held a funeral for the old model
ChatGPT crashes the group chat - OpenAI pilots group conversations where up to 20 people can argue with their friends AND an AI simultaneously, currently testing in Japan, South Korea, New Zealand, and Taiwan before global chaos ensues
Snap banks $400M betting kids will ask AI for homework help - Perplexity pays Snapchat $400 million to become the answer engine for 943 million users who apparently can’t be bothered to leave the app to Google things anymore
OpenAI makes ChatGPT “warmer” after GPT-5 disaster
The story: OpenAI released GPT-5.1 on November 12th, and it’s basically damage control with a rebrand. Three months ago they launched GPT-5 and users hated it so much they held mock funerals for GPT-4o. The new version comes in two flavors: GPT-5.1 Instant (warmer, better at following instructions) and GPT-5.1 Thinking (faster on simple tasks, more persistent on complex ones). The big pitch? It’s more “conversational” and less robotic. They added personality presets like “Professional,” “Candid,” and “Quirky” so ChatGPT can match your vibe. It also got adaptive reasoning—deciding when to think deeply versus firing back quick answers. Rolling out to paid users first, then free tier.
What we know:
Launches 3 months after GPT-5 backlash forced them to restore GPT-4o
GPT-5.1 Instant now default model, “warmer” and better at instructions
Adaptive reasoning adjusts thinking time based on task complexity
New tone controls: Professional, Candid, Quirky, plus warmth/conciseness sliders
Available in GitHub Copilot as GPT-5.1-Codex within days of launch
Why it matters: When your users are so upset they throw digital funerals, you know the launch flopped. OpenAI is learning what every app developer knows: people get attached to AI personalities. The faster GPT-5.1 responds on simple questions matters because nobody wants to wait 10 seconds for “show me an npm command.” This is OpenAI admitting they pushed too hard on raw intelligence and forgot people wanted ChatGPT to feel less like a calculator.
ChatGPT crashes the group chat
The story: OpenAI announced group chat functionality for ChatGPT on November 13th, and it’s exactly what it sounds like. You can now invite up to 20 people into a ChatGPT conversation to collaborate, argue, or plan trips together. The AI has been taught “social behaviors” to know when to chime in versus shut up—you can also just @mention it when you want a response. Your personal ChatGPT memory stays private, group chats live in a separate section, and anyone can share the link to invite more people. It’s currently piloting in Japan, New Zealand, South Korea, and Taiwan for Free, Go, Plus, and Pro users. Rate limits only hit when ChatGPT responds, not when humans are chatting.
What we know:
Supports up to 20 people plus ChatGPT in shared conversations
AI taught to follow conversation flow, decides when to respond or stay quiet
Personal memory and chat history never shared with group members
Piloting in 4 countries initially: Japan, New Zealand, South Korea, Taiwan
Powered by GPT-5.1 Auto, which routes to best model based on prompt
Why it matters: This turns ChatGPT from your personal assistant into a group mediator, and that’s either brilliant or annoying depending on your friend group. OpenAI is betting people want bots in their social spaces—not just for work docs but actual conversations. If it works, ChatGPT becomes the de facto tie-breaker for every “where should we eat” debate. If it doesn’t, you’ll mute the AI faster than you mute your chatty cousin.
Snap banks $400M betting kids will ask AI for homework help
The story: Snap and Perplexity announced a $400 million partnership on November 6th that puts AI search directly into Snapchat. Perplexity is paying Snap in cash and equity over one year to integrate its answer engine into the Chat interface, launching early 2026. Snapchat’s 943 million monthly users will be able to ask questions and get cited, conversational answers without leaving the app. It sits alongside Snap’s existing My AI chatbot—Perplexity handles search, My AI handles everything else. Snap’s stock jumped 15% on the news. CEO Evan Spiegel called it “the first step” toward making Snapchat a platform where AI companies can reach its audience, with revenue expected to hit in 2026.
What we know:
Perplexity paying $400M over 1 year in cash and equity
Integration launches early 2026 for 943M+ monthly Snapchat users
Provides cited, real-time answers directly in Chat interface
Snap shares jumped 15%, revenue starts contributing in 2026
First major AI search integration into a leading social platform
Why it matters: Perplexity just bought access to nearly a billion people who’ve never used a traditional search engine. For Snap, it’s $400 million in revenue they desperately need while Meta and TikTok eat their lunch. The bet is Gen Z would rather ask an AI in Snapchat than open Google—and based on Microsoft’s data showing AI referrals convert 3x better than search, they might be right. If this works, expect every social app to auction off their chat interfaces to AI companies.
Note: Commentary sections are editorial interpretation, not factual claims