Wake up, check your phone, sip coffee — and ChatGPT has already done a little homework for you. On September 25, 2025, OpenAI rolled out ChatGPT Pulse, a mobile-first feature that quietly runs overnight, combs your past chats and any connected apps you opt into (think calendar, email, contacts), and delivers a handful of curated, scannable cards each morning: quick race results, a suggested warm-up and menu tips for tonight’s dinner, a short vocabulary drill for the language you’re learning — whatever the model decides is useful for your day. The company is positioning Pulse as less of a newsfeed and more of a compact, proactive assistant nudge.

What Pulse actually does

Pulse is not a new chat window. It’s a one-a-day research assistant: the system runs “asynchronous research” on your behalf while you sleep and surfaces 5–10 short updates in the form of visual cards you can either scan quickly or tap for details. OpenAI describes it as a way to start each morning “with a new, focused set of updates” tailored to you — built from your chat history, explicit feedback about what you like, and any apps you link. The experience is currently available as a preview on mobile and limited to ChatGPT Pro subscribers while OpenAI irons out scale and safety trade-offs.

A demo that reads like a planner

Reporters who saw Pulse in action walked away with a surprisingly tangible sense of how hands-on it can get. In a demo reported by The Verge, personalized prompts referenced a user’s calendar, dietary preferences and recent chat threads: Pulse suggested a 45–50 minute running route that conveniently ended near the rooftop restaurant they were booked for, offered a “dinner strategy” tailored to a dairy-free diet, and even proposed a post-run buffer in case plans ran late. It also asked for short feedback prompts like “What’s on your mind lately?” to shape tomorrow’s briefing — so the system can tune itself to what you actually care about without you writing long instructions.

The data trade: opt-in, but intimate

OpenAI is clear that Pulse only gets access to the sources you explicitly allow. If you already have “reference past history” turned on, Pulse will use your prior chat transcripts; if you connect your calendar or email, you’ll be asked to grant permission for those apps to be read as well. The company says the product is configurable — you decide which apps are in play and you can give feedback to shape the kinds of updates you receive. That said, the feature’s usefulness is proportional to how much context you let it access, and that raises obvious privacy and safety questions.

Privacy and safety: the concern everyone mentions

Pulse’s design intentionally avoids the infinite scroll; OpenAI product leads told journalists the experience “ends” rather than aims to maximize attention. The company also says Pulse won’t use your personalized briefing to train models for other users, and that the feedback you give will improve your Pulse, not the global model. Still, critics flag the core risk: an assistant that learns your habits and preferences is also capable of reinforcing patterns — desirable and undesirable — and might overfit to a narrow view of what you want to see. OpenAI says it has safety filters and that its policy teams are actively evaluating the mental-health and echo-chamber implications, but specifics beyond high-level assurances are thin in the early preview.

Why Pro first? Cost, compute and control

OpenAI is launching Pulse behind its Pro paywall for a reason: the feature is compute-heavy. Building proactive, personalized overnight research for millions of users is expensive, so the company is testing with Pro subscribers before expanding more broadly. That mirrors a pattern we’ve seen across Big Tech: new, agent-like capabilities tend to debut in premium tiers before wider rollout, because the economics and moderation requirements are still being worked through. Analysts also frame Pulse as a stepping stone toward fuller “agents” — systems that not only research but take action on your behalf.

The bigger game: from reactive chatbot to ambient assistant

Pulse is less a product than a signal of intent. OpenAI and its rivals increasingly talk about building AI that doesn’t wait to be asked — assistants that understand goals, anticipate needs and take steps toward outcomes. Fidji Simo, OpenAI’s head of applications, and other executives have framed that transition as the next frontier: the shift from reactive Q&A to background orchestration and action. Pulse is an early, contained version of that vision — helpful in day-to-day planning, but also a preview of higher-stakes design decisions to come (automation, delegation, who gets to act for you).

What it means for you

If you live in an ecosystem that values convenience over friction, Pulse will look like an elegant way to make mornings more efficient. If you’re privacy-conscious, the idea of a proactive assistant reading your email and calendar overnight will feel invasive, even with opt-ins and promises. Either way, Pulse forces a conversation about what “help” from AI should look like: nudges and logistics, or active intervention and decision-making. For now, OpenAI is betting people will try the convenience first and reckon with the tradeoffs later.

Pulse is small in one sense — a daily set of cards — and huge in another: it’s the productization of an idea long teased by AI labs everywhere. Whether users embrace a morning brief written by a neural net will depend on how well OpenAI balances utility, transparency and safety as the feature moves out of Pro preview and into the broader world.