OpenAI drops ChatGPT Pulse to research before you even ask

ChatGPT is tired of waiting for you to type.

OpenAI just rolled out ChatGPT Pulse, a new mobile preview that works the night shift while you sleep, delivering a personalized briefing of five to ten visual cards to kick off your morning.

The feature is currently exclusive to Pro subscribers paying $200 a month, with plans to expand to Plus members before eventually reaching everyone.

Your morning brief, built overnight

Since its launch, ChatGPT has been fundamentally reactive: you prompt, it answers. Pulse completely flips that dynamic.

Running asynchronously each night, Pulse synthesizes your chat history, memory, direct feedback, and connected apps like Gmail and Google Calendar to anticipate what you need. When you wake up, you get interactive visual cards with AI-generated text and images covering anything from sports roundups to actionable plans.

During a demo with The Verge, Christina Kaplan, who leads personalization and productivity for ChatGPT, showed how Pulse mapped out a 45-minute running route that finished near a work dinner venue, complete with menu recommendations tailored to her dairy-free diet.

OpenAI product lead Adam Fry showed off cards tracking Arsenal soccer news, family Halloween costume ideas, and a toddler-friendly itinerary for an upcoming trip to Sedona. If you want specific topics, you can tap curate to request updates on professional tennis or local weekend events, or steer future editions with simple thumbs-up and thumbs-down reactions.

Crucially, OpenAI opted out of doomscrolling. After a handful of reports, the feed stops and displays "that's it for today."

Agent dreams and the $200 paywall

Pulse is deeply resource-heavy. CEO Sam Altman previously teased that compute-intensive features would land on the Pro tier first, and Fry admitted the compute spent on Pulse can vary tremendously depending on the web research required.

To make it actually useful, users have to hand over keys to their digital lives. Integrations with Gmail and Google Calendar are off by default, but taking advantage of proactive scheduling requires explicit consent to parse your daily schedule and messages. OpenAI notes that user feedback only trains personal Pulse instances, not global models.

According to Fidji Simo, OpenAI's CEO of applications, most current agents still wait around to be told what to do. The goal with Pulse is to lay the groundwork for software that grasps long-term goals and acts autonomously before you even think to ask.

Why it matters

We think Pulse marks a pivotal shift from the reactive search-box era of AI toward true personal agency.

For users, the value lies in eliminating the mental overhead of deciding what to prompt. Having an assistant that connects calendar gaps, dietary needs, and research topics into tactical morning steps is genuinely compelling. But at $200 a month, it is impossible to recommend right now for anyone other than enterprise power users or curious tech elites. Everyday users should wait until it trickles down to Plus.

For the broader industry, Pulse puts traditional morning newsletters, news aggregators like Apple News, and workflow reminders on high alert. When an assistant possesses full conversational context alongside calendar data, generic content curation looks increasingly dated.

OpenAI's decision to cap the feed with a hard stop is a remarkably clever design choice that respects user attention. However, proactive suggestions walk a razor-thin line between magical convenience and privacy intrusion. If overnight research misses the mark or pulls uncomfortable assumptions from sensitive emails, user trust could evaporate fast.