OpenAI has introduced a new feature called Pulse within ChatGPT, designed to generate personalized reports while users sleep. Pulse delivers five to ten briefings to help users quickly catch up on the day’s developments, aiming to encourage users to check ChatGPT first thing in the morning, much like they would social media or news apps. Pulse is part of a broader shift in OpenAI's consumer products, which have recently been designed to serve users asynchronously rather than simply answering queries. Features like ChatGPT Agent or Codex aim to make ChatGPT more of an assistant than a chatbot. With Pulse, OpenAI appears to be steering ChatGPT toward a more proactive role. “We are building an AI that enables us to offer support previously accessible only to the wealthiest individuals, and gradually make it available to everyone,” stated Fidji Simo, the new CEO of OpenAI’s consumer applications, in a blog post. “ChatGPT Pulse is our first step in this direction—starting with professional users today, but with the goal of democratizing access to this intelligence.” Sam Altman, CEO of OpenAI, mentioned earlier this week that some of ChatGPT’s “computationally intensive” features will be limited to the company’s most expensive subscription tiers—Pulse falls into this category. OpenAI has previously stated that it faces significant constraints in the number of servers powering ChatGPT and is actively building AI data centers in collaboration with partners like Oracle and SoftBank to expand its capacity. Starting Thursday, OpenAI will roll out Pulse to subscribers of its $200-per-month Professional Plan, who will see a new tab in the ChatGPT app. The company plans to eventually extend Pulse to all ChatGPT users, with Plus subscribers gaining access soon, pending improvements in the product’s efficiency. The reports generated by Pulse can include news summaries on specific topics—such as updates on a favorite sports team—as well as more personalized briefings based on user context. In a demo, Adam Fry, OpenAI’s product lead, showcased several Pulse reports generated for him: a summary of news related to the UK football team Arsenal; group Halloween costume ideas for his wife and children; and a toddler-friendly itinerary for a family trip to Sedona, Arizona. Each report appears as a “card” containing AI-generated images and text. Users can click on each card for the full report and then ask ChatGPT follow-up questions. While Pulse automatically generates reports, users can also request new ones or provide feedback on existing ones. A key aspect of Pulse is that it stops generating reports after delivering a few and displays a message: “Great, that’s all for today.” According to Fry, this is a deliberate design choice to differentiate the service from social media apps, which are optimized for engagement. Pulse integrates with ChatGPT Connectors, allowing users to link apps like Google Calendar and Gmail. Once set up, Pulse will parse your emails overnight to highlight important messages in the morning or access your calendar to create agendas for upcoming events. If users have enabled ChatGPT’s memory feature, Pulse will pull context from past conversations to enhance report personalization. Christina Wadsworth Kaplan, OpenAI’s head of personalization, gave an example of how Pulse automatically recognized her passion for running and created a London trip itinerary that included running routes. Wadsworth Kaplan described Pulse as a “new kind of feature” for consumer products. As someone who follows a pescatarian diet, she noted that Pulse reviewed her dinner reservations on her calendar and found menu options that matched her dietary preferences. However, it's hard to ignore how Pulse competes with existing news products like Apple News, paid newsletters, or traditional media. Fry doesn’t expect Pulse to replace the various news apps people use, noting that the feature, like ChatGPT Search, cites its sources and provides links. Whether Pulse justifies the computational resources it requires remains to be seen. Fry explained that the computing power needed for the service can vary significantly depending on the task—some items are quite efficient, while others may involve searching the web and synthesizing large volumes of documents. Ultimately, OpenAI envisions making Pulse more agent-like, capable of booking restaurants or drafting emails for user approval. However, such capabilities are likely still far off and may require significant advancements in OpenAI’s agent models before users feel comfortable delegating such decisions to AI.