Table of Contents
- The Twelve-Month Timeline
- The Shared Bet
- Where Each Product Actually Differs
- What Three Labs Converging Tells You
- The Three Things That Still Have to Be Solved
- How to Choose Between Them
- FAQ
Look at the calendar. OpenAI shipped ChatGPT Pulse in September 2025. Anthropic shipped Orbit on May 6, 2026, at Code with Claude. Google previewed Gemini Proactive Assistance at I/O on May 19. In a twelve-month window, the three largest AI labs in the world all bet on the same product — a proactive assistant that reaches out before you ask, generates daily briefings from your work tools, and surfaces actions across apps without waiting for a prompt.
When three labs that disagree about almost everything else converge on the same shape, the pattern matters more than any individual product. The interesting question is not which proactive assistant wins. It is what the convergence is telling you about where the AI surface is going.
The Twelve-Month Timeline
September 2025 — ChatGPT Pulse. OpenAI shipped Pulse as its first proactive, asynchronous assistant. Overnight briefings generated from chats, memory, and connected Gmail and Calendar. The product set the template: a personalized daily summary delivered without the user opening the app.
May 6, 2026 — Anthropic Orbit. Anthropic debuted Orbit at its Code with Claude developer conference. Two halves: a proactive briefing assistant for Claude Cowork that connects to Gmail, Slack, GitHub, Calendar, Drive, and Figma, and a separate mobile agent that operates iOS and Android. The briefing half is the direct Pulse competitor; the phone agent is its own category.
May 19, 2026 — Google Gemini Proactive Assistance. Previewed at I/O. An agentic system designed to operate across apps and browsing sessions without waiting for a user prompt. Pulls activity data across connected services. Integrated with Android, ChromeOS, and the Gemini app.
Three labs. Three months apart in actual shipping, but functionally the same product class hitting the market inside a year.
The Shared Bet
All three products are betting on the same shift in user behavior: that people will stop opening assistants and start receiving them. The chat-as-destination model treats the assistant as a place you go when you have a question. The proactive model treats the assistant as a layer that watches what is happening in your work tools and surfaces what matters without you asking.
That shift maps onto a bigger pattern across the AI surface. Cursor 3 demoted the IDE. Claude Design pulled visual creation into the chat. Now proactive AI is pulling the user out of the chat entirely. The assistant comes to you in whatever surface you already use — a notification, an email, an inline suggestion — rather than waiting for you to open a chat tab.
The structural bet is that the dominant interaction model for AI in 2027 is not the chatbot. It is ambient. The chat surface becomes the configuration UI for an assistant that mostly operates without the chat being open. All three labs reading this trend at the same time and shipping similar products at the same time is the strongest signal anyone is going to give you that the bet is real.
Where Each Product Actually Differs
The shared shape obscures real differences in positioning. Each lab is targeting a different user.
ChatGPT Pulse targets the executive. Connector set is email- and calendar-heavy. The default briefing reads like a personal chief of staff drafted it. The implicit user has too many inboxes and not enough time to read all of them.
Anthropic Orbit targets the builder. Connector set includes GitHub and Figma alongside the standard work tools. The default briefing reads like a tech lead drafted it — what changed in your repos, what’s blocked in your design files, what your team committed since you signed off. The implicit user is a developer, designer, or product manager whose work lives in version-controlled artifacts.
Google Gemini Proactive Assistance targets the consumer. The product is integrated with Android, ChromeOS, and the Gemini app surface that already has a billion daily-active users. The implicit user is anyone who uses Google’s existing services and would benefit from an assistant that operates across them without being asked.
The connector sets and default surfaces tell you who each lab thinks the durable proactive-AI user is. None of them are competing for the same user yet. Six months from now, they will be.
What Three Labs Converging Tells You
Three structural conclusions land harder because three labs all bet on them.
The chat interface is no longer the primary AI surface. Every lab now has an architecture where the chat is a configuration and review tool, not the place work happens. Deep Research showed it. Cursor’s Agents Window showed it. Orbit and Pulse and Gemini Proactive show it for the consumer surface. The chat is necessary but not sufficient.
Connector breadth is now table stakes. All three products live or die on how many of your work tools they can read. The competitive question is not “does the AI work” — it is “does it work against my actual tool stack.” Expect the connector race to dominate AI product roadmaps through Q3.
The frontier-lab model now includes the assistant. A year ago, frontier labs sold APIs to developers who built assistants. Now all three frontier labs sell the assistant directly. The platform-versus-product line has moved, and the labs are on the product side of it. Apple’s Gemini-Siri deal tells you what happens when a platform company recognizes the labs are now the consumer surface.
The Three Things That Still Have to Be Solved
Three open problems show up in all three products and none have been solved by any of them.
Information overload. A proactive assistant that surfaces too much becomes a notification firehose. Pulse, Orbit, and Gemini Proactive Assistance all have to win the relevance problem — knowing what matters versus what merely changed. The defaults are early. The personalization is unproven at scale.
Trust calibration. Users have to learn when to trust the assistant’s surfacing and when to verify. Calibration takes weeks of use and breaks the moment the assistant gets a high-stakes call wrong. None of the three products have shipped the trust-calibration UX cleanly yet.
Privacy disclosure. Reading Gmail, Slack, and your design files in the background creates a data-flow disclosure problem all three products are handling at varying levels of clarity. Expect regulatory attention in the second half of 2026, especially in Europe.
How to Choose Between Them
Pick by where you already work, not by which lab you prefer.
If your day lives in email, calendar, and chat, and you do not write or design for a living, Pulse is the strongest fit. The connector set is built for you and the briefing voice maps to executive-style work.
If your work artifacts are in repositories, design files, or work-management tools alongside email, Orbit’s briefing assistant targets you specifically. GitHub and Figma in the connector list is not decoration — it is the differentiator.
If your stack is Google-centric and you are willing to wait for full rollout, Gemini Proactive Assistance has the deepest integration with Android, ChromeOS, and Workspace. The launch is in preview; the full version arrives over the rest of 2026.
Do this first: pick the proactive assistant that matches the tool stack you actually use, run it for two weeks at the default settings, then audit what it surfaced. Read every briefing it generated. The pattern of what it considered worth surfacing tells you whether the relevance model fits your work — and if it does not, no amount of tuning will fix it. Move to the next product or wait six months for the defaults to mature.
FAQ
What is a proactive AI assistant?
An assistant that reaches out before you ask. Rather than waiting for a chat prompt, it watches your connected work tools, generates briefings, and surfaces actions across apps. ChatGPT Pulse, Anthropic Orbit, and Google Gemini Proactive Assistance are the three current entries in the category, all shipped within a twelve-month window.
Which proactive assistant is best?
Depends on your tool stack. Pulse targets email-and-calendar executives. Orbit targets builders with GitHub and Figma in their workflow. Gemini Proactive Assistance targets Google-ecosystem users. None of the three are clearly better — they are clearly aimed at different users.
Are these the same as voice assistants like Siri?
No. Voice assistants are reactive — you ask, they answer. Proactive assistants generate output without being asked. The new Gemini-powered Siri is moving in the proactive direction, but the core category here is text-and-action briefings rather than voice interaction.
Do they read my email and Slack?
Yes, with permission. All three connect to email, calendar, and chat platforms (Pulse mostly Gmail and Calendar, Orbit includes Slack and more, Gemini integrates with Workspace). The connection is opt-in and revocable, but the read access is necessary for the products to work. Privacy disclosure varies; check each product’s terms before connecting sensitive accounts.
Can I use more than one?
Yes. Nothing prevents running Pulse for executive email triage and Orbit for builder-tool briefings, for example. The cost is divided attention to two daily briefings rather than one. Most users will eventually consolidate around the assistant whose connector set best matches their daily work.
Is proactive AI the future of assistants?
Three frontier labs shipping the same shape in twelve months is the strongest signal anyone is going to give that the answer is yes. The chat interface remains for configuration and review, but the default interaction is moving from “open the assistant and ask” to “the assistant reaches out when something matters.”