ChatGPT Pulse and the missing primitives

When OpenAI launched ChatGPT Pulse in September 2025, Fidji Simo framed it as taking "the level of support that only the wealthiest have been able to afford" and making it available to everyone. I was genuinely excited when I heard that. What actually shipped is a morning news digest that knows you pretty well.

Pro subscribers open the app each morning, scan 5–10 personalized cards, and occasionally find something they wouldn't have discovered on their own. The early reviews are consistent: it's good, it's useful, people would notice if it disappeared. But it doesn't quite feel like the proactive assistant OpenAI described.

I've been thinking about why, and honestly it comes down to infrastructure. The three-primitives framework makes the gaps pretty clear.

One primitive filled in, two still empty.

What Pulse gets right

Credit where it's due, the clock works.

Every night, Pulse processes your chat history, connected apps (Gmail, Google Calendar), and explicit feedback to generate personalized morning cards. The personalization is genuinely impressive. Users report cards referencing offhand mentions from weeks-old conversations, connecting dots they didn't draw themselves. One Reddit user described a card about a birthday they'd never explicitly mentioned — Pulse inferred it from calendar access.

The format is smart too. Cards are visual, scannable, and intentionally capped. "That's it for today" is a design choice that respects attention rather than trying to maximize engagement time. That restraint separates Pulse from a social media feed, at least in intent.

And the connected-app integration adds real context. When Pulse can see your calendar and email, the morning briefing gains a layer of usefulness that pure chat-history analysis can't match. It knows you have a meeting at 10am and surfaces context about the person you're meeting with.

All of this works because the clock primitive is well-built. The overnight schedule runs reliably, the personalization engine has enough signal to work with, and the delivery format is thoughtful.

What useful → indispensable would look like.

What's missing

So the clock runs at night. Everything else waits until morning.

If a customer replies to a critical email at 2pm, Pulse tells you about it in tomorrow's digest. If a deploy fails at 11am, you find out the next day (assuming Pulse even has access to your deploy pipeline, which it doesn't). If a competitor ships a new feature at noon, same story: tomorrow.

This is the listener gap. Pulse has no real-time change detection from external systems. It reads your email history overnight, but it doesn't watch your email for urgent changes. It can see your calendar, but it can't push a notification when someone reschedules a meeting five minutes from now. It processes a snapshot of the world once per day and tells you what it found.

For a morning briefing, that's fine. For the proactive assistant OpenAI described in the announcement, it's not enough.

The inbox gap is subtler. Pulse delivers results in one direction: cards in the ChatGPT app that you consume passively. You can give thumbs up or thumbs down, but you can't tell Pulse to deliver a specific result to Slack, or file a ticket, or send a draft email. The delivery channel is fixed. A proactive assistant that can only talk to you through morning cards is like an intern who can only communicate via Post-it notes left on your desk overnight.

The SentiSight analysis called it "incremental evolution rather than revolution" and compared it to Google Now from 2012. That comparison stings, but it's not unfair. The architecture is structurally similar. A scheduled job processes your data overnight and surfaces what it thinks matters. The AI is dramatically better than 2012. The architecture isn't.

The reception confirms the gap

The early reactions to Pulse tell a consistent story.

The Hacker News discussion was skeptical. Users compared it to algorithmic feeds, worried about privacy, and called out the underlying business model. One commenter pointed to OpenAI's projected $25 billion in revenue from free users by 2029, suggesting Pulse is ultimately an advertising delivery mechanism. Multiple commenters referenced dystopian fiction: Bradbury, Ted Chiang, Ken Liu. The concern wasn't that Pulse is bad. It's that proactive AI pushing content at you is a slippery slope toward manipulation.

Matthew Becher's Substack review was more positive. He credited Pulse with genuinely nudging him forward on side projects. But he asked the right question: "Will proactive AI feeds make people smarter, more curious, and more productive, or will passivity make it just another feed to mindlessly scroll through?" After a few weeks, he found the suggestions becoming repetitive and requiring active redirection to stay useful.

The Reddit reactions were impressed by the personalization but unsettled by the privacy implications. The birthday-inference example mentioned above was described as "helpful and creepy in equal measure," a phrase that captures the general mood.

What I didn't see anyone say was "this changed how I work." The responses mostly landed somewhere between "interesting" and "a little unsettling."

That said, this is a curated sample. The replies under Sam Altman's announcement post on X were overwhelmingly positive. Many Pro users report Pulse as the feature that justifies the subscription. For a consumer product, "useful morning briefing that people open every day" may be exactly the right scope. The gap we're describing below is specific to people who want agents that act, not just inform.

What would make it indispensable

Imagine a version of Pulse with all three primitives.

The clock still runs overnight, generating your morning briefing. But it also runs in real-time. When something changes in a system you care about, you find out immediately, not tomorrow.

A listener monitors your connected systems for meaningful changes: a customer reply that's been waiting more than four hours, a deploy that failed, a PR that's been in review for three days, a competitor's changelog that just updated. Not everything. Just the things you told it to care about, delivered as structured change events with context about what moved and what it was before.

An inbox delivers results wherever they're useful. A morning card when a digest is the right format. A Slack DM when something is urgent. A draft email when the right action is a reply. A filed ticket when the right action is tracking. The delivery channel matches the action, not the product's UI constraints.

That version of Pulse would actually look like the proactive assistant OpenAI described in the announcement. The models are good enough to handle every judgment call on this list, and honestly they have been for a while. What's missing is the infrastructure underneath, and that's the harder part to build.

The market signal

The most interesting thing about Pulse might be the competitive landscape around it. Anthropic is reportedly building a proactive assistant called Orbit for Claude. Google and Perplexity are developing their own versions. Everyone is converging on the same insight: reactive AI (ask a question, get an answer) is leaving value on the table.

That's exciting and it validates what we've been building. But if everyone just ships the clock and calls it done, I'm not sure any of these will feel like more than a tab in an app.

The pattern in the reviews is consistent: people open Pulse, find something useful, and close the tab. A proactive assistant that could push a Slack message when a deploy breaks or file a ticket when Sentry catches a new error class would be fundamentally different — the kind of tool you'd reorganize your workflow around, not just check over coffee.

Evaluating any proactive product

The Pulse analysis above follows a pattern that works for any product claiming to be proactive. Three questions, scored yes or no:

Primitive	Question	What "yes" looks like
Clock	Can it wake itself up on a schedule or interval?	Cron, timer, overnight batch. Most products have this.
Listener	Can it react to external changes in real time?	Webhooks, streaming APIs, change feeds. Not polling on a timer.
Inbox	Can it deliver results to different channels based on context?	Slack for urgent, email for digest, ticket for tracking. Not locked to one UI.

Most products today score one out of three because the clock is the easiest to build. Getting all three working is where it starts to actually feel like the proactive assistant everyone keeps promising.