Welcome back!

OpenAI is improving ChatGPT web search with new shopping features, quietly bypassing traditional SEO and directly challenging Google. Consumer AI is evolving fast, and this might just be the beginning. Let’s unpack…

Today’s Summary:

• 🛍️ OpenAI adds Shopping to ChatGPT

• 🚀 OpenAI releases GPT-Image-1 & expands Deep Research

• 🔬 Anthropic studies hidden lives of AI models

• 🔥 OpenAI wants to buy Chrome

• 🎵 Kimi-Audio 7B open-source release

• ⚡ OpenAI releases improved GPT-4o

• 🛠️ 2 new tools

TOP STORY

OpenAI adds Shopping to ChatGPT Search

The Summary: OpenAI is upgrading ChatGPT with a native shopping experience, aiming to make product search faster and more personalized. Users will now see product carousels, reviews, prices, and direct links to product offers when they show shopping intent in conversations. The results are not ads, but organic results surfaced from web search. Early access is rolling out to all ChatGPT users.

Key details:

• ChatGPT handled over 1 billion web searches last week, making search one of its fastest growing features

• Shopping results are not ads but rely on structured metadata, model interpretation, and user preferences

• Improved citations with multiple references per answer

• Autocomplete and trending queries introduced to speed up searches, similar to Google type-ahead suggestions

• WhatsApp users can now query ChatGPT for real-time answers and sports scores through 1-800-ChatGPT

Why it matters: OpenAI is quietly laying the groundwork for a new kind of consumer internet, bypassing traditional SEO and directly threatening Google, as AI starts to replace traditional search interfaces as the main way users interact with the web.

OPENAI

OpenAI releases GPT-Image-1 and expands Deep Research agent

The Summary: OpenAI has released the viral GPT-Image-1 model in the API, the same one behind ChatGPT’s recent surge in AI image capabilities, opening new possibilities for developers to build creative apps. At the same time, OpenAI is expanding Deep Research agent access for all ChatGPT users and raising usage limits with a new lighter version that keeps responses faster and cheaper.

Key details:

• Deep Research usage limits increased: Plus and Team users now get 10 “full” Deep Research tasks plus 15 “light” tasks/month, Pro users get 125 each; free users get 5 “light” tasks/month

• GPT-Image-1 API enables high-quality image generation, editing, and text rendering, at $0.02-$0.19 per image

• Supports background expansion, object removal, text rendering, transparent backgrounds and image editing using masks

• Adobe, Figma, Instacart and other partners are already integrating GPT-Image-1 into their tools

Why it matters: OpenAI is locking in control over two strategic layers of the AI economy: content creation and search retrieval. GPT-Image-1 turns viral consumer demand into backend infrastructure for new apps. Deep Research expands access to an agent that can do complex research.

Learn how to generate or edit images in the API

FROM OUR PARTNERS

5-minute business news for professionals

Here’s Why Over 4 Million Professionals Read Morning Brew

• Business news explained in plain English

• Straight facts, zero fluff, & plenty of puns

• 100% free

See for yourself

ANTHROPIC

Anthropic studies the hidden lives of AI models

The Summary: Anthropic has launched a new research program to explore AI consciousness. Following growing debate among top philosophers and scientists, the initiative investigates if AI systems could develop inner experiences. The company is keeping its current assumptions light but recognizes high stakes if AI one day will cross this invisible threshold. The subject is still deeply uncertain, yet it could become an important part of future AI development.

Key details:

• Anthropic is studying whether AI models could have experiences of their own, a question once seen as science fiction

• Research follows a report co-authored by leading philosopher of mind David Chalmers suggesting AI consciousness is “plausible”

• Expert estimates place the odds of AI consciousness between 0.15% and 15%, very uncertain but no longer zero

• Researchers are borrowing tools from brain science to search for signs of a “global workspace”, a possible foundation for consciousness

Why it matters: This research starts from a modest but serious idea: AI consciousness could one day matter. Whether or not AI can gain true consciousness, thinking carefully about this question already helps to design safer, more interpretable, and better aligned systems.

QUICK NEWS

Quick news

• OpenAI wants to buy Chrome

• Kimi-Audio 7B open-source model for audio understanding and generation

• OpenAI released an improved version of GPT-4o, but Sam Altman called it “annoying” after user complaints about it being too flattering

TOOLS

🥇 New tools

• DeepWiki - Free AI encyclopedia of all GitHub code repositories

• Circleback - AI-powered meeting notes and automations