OpenAI Spaces and Pages: ChatGPT's Answer to Microsoft Copilot
OpenAI's Spaces and Pages bring shared, editable documents into ChatGPT, letting teams and AI agents work together in one thread.
What are Spaces and Pages in ChatGPT?
Spaces is OpenAI’s new shared workspace inside ChatGPT where people and AI agents work on the same material together, instead of each person running their own private chat thread. Pages are the documents that live inside those Spaces: editable files that both humans and the AI can revise, building on the same context thread rather than starting a conversation from scratch each time. Announced at DevDay alongside the Dots agent and GPT-5.1 Codex, the pairing puts OpenAI squarely into the collaborative document territory that Microsoft 365 and Google Workspace have owned for years.
TL;DR
- Spaces gives teams a shared place to work with ChatGPT instead of siloed, one-off chats, so context from a project persists across a group rather than getting trapped in a single person’s history.
- Pages are the editable documents inside a Space, meaning a team and an AI agent can co-author material and keep revising it as a project evolves, rather than exporting chat output into a separate doc.
- This is a direct move into Microsoft’s territory: Spaces and Pages look like an early version of a Word or Teams competitor built natively around an AI that already has your context.
- OpenAI paired Spaces/Pages with Dots, an agent with persistent memory and its own cloud computer, suggesting the real bet is that accumulated context in ChatGPT becomes more valuable the longer a team stays inside the ecosystem.
- A cheaper, more efficient model (GPT-5.1 Codex, referred to in the source material as Soul/Sol 6.1) makes it economically realistic to run more collaborative, agent-assisted work without blowing through usage limits.
- The bigger strategic signal is that OpenAI wants more of your team’s work to never leave ChatGPT, which feeds the same context back into future agent tasks.
- For builders, the opportunity is less about using Spaces personally and more about the application layer opening up around agents that operate inside ChatGPT.
Built like a system. Not vibe-coded.
Remy manages the project — every layer architected, not stitched together at the last second.
Why would OpenAI build a Microsoft Office competitor?
Because the real asset in enterprise AI isn’t the model, it’s the context the model has about your work. OpenAI has already built a massive consumer base for ChatGPT (OpenAI has cited over a billion users at DevDay), and a huge share of ongoing work that could generate reusable context: drafts, decisions, project history, half-finished documents. Right now, a lot of that work still lives in Google Docs, Slack threads, or Microsoft Word, which means ChatGPT only sees fragments of it when someone pastes in a question.
Spaces and Pages change the default location of the work itself. If a team drafts a launch plan, a sales deck, or a product spec as a Page inside a Space, the AI isn’t reconstructing context from a copy-pasted snippet. It already has the full document, the edit history, and the running conversation around it. That’s a fundamentally different starting position than a chatbot answering questions about a file you uploaded.
This is also why the comparison to Microsoft 365 Copilot is apt rather than hyperbolic. Microsoft’s pitch has always been “AI inside the documents you already use.” OpenAI’s pitch with Spaces and Pages flips that: “documents inside the AI you already use.” Whoever controls where the actual work product lives controls the richest, most current context, and context is what makes an agent genuinely useful rather than generically smart.
How does this connect to Dots and the new agent push?
Spaces and Pages don’t exist in isolation. They launched alongside Dots, an OpenAI agent that runs on the company’s top model, keeps its own memory, operates its own cloud computer, and can keep working after a person logs off. Dots is reachable from the ChatGPT web app, mobile, Slack, and Codex, and it’s built to take on recurring responsibilities rather than answer one-off prompts.
The connection to Spaces is direct: a shared workspace full of live documents is exactly the kind of environment a persistent agent needs to be useful over time. If a product launch date slips, a Dot with access to the relevant Space can, in principle, catch which sections of a sales deck, launch page, or Page document depended on that date and prepare the necessary revisions proactively. That’s a materially different value proposition than an assistant that only responds when asked.
Notably, Dots only draws on memories explicitly shared through ChatGPT’s memory settings or through connected applications you authorize, rather than ingesting everything by default. That’s a deliberate privacy boundary: users decide what context an agent can see, which matters a lot once agents are embedded in shared team workspaces with sensitive documents.
Is this actually a threat to Microsoft Copilot?
Other agents ship a demo. Remy ships an app.
Real backend. Real database. Real auth. Real plumbing. Remy has it all.
It’s early, but the structural threat is real, even if the feature set isn’t fully mature yet. Microsoft’s advantage has always been distribution: Word, Excel, Teams, and Outlook are already installed everywhere enterprises work, and Copilot rides on top of that installed base. OpenAI doesn’t have that distribution inside the enterprise document stack. What it has instead is a product people already trust for drafting, reasoning, and research, and a growing habit of doing real work directly inside ChatGPT rather than treating it as a lookup tool.
Spaces and Pages are OpenAI’s attempt to convert that habit into a genuine workplace surface, not just a chat window. If teams start drafting specs, running project discussions, and iterating on documents inside Spaces, OpenAI starts capturing the same kind of operational context that currently lives in Microsoft’s ecosystem. That’s the long game: not winning on feature parity with Word on day one, but making ChatGPT the place where work originates, so the context advantage compounds.
The open question is adoption friction. Enterprises have deep, sticky investments in Microsoft’s suite, including permissions systems, compliance tooling, and integration with existing IT infrastructure. Spaces and Pages will need to prove they can slot into that environment, or give teams a strong enough reason to route around it, before this becomes a real share shift rather than a parallel option for AI-forward teams.
What should builders and teams actually do with this?
Treat Spaces as a place to consolidate recurring, high-context work rather than a replacement for every document you own. The clearest value shows up when a team is iterating on something continuously (a product spec, a launch plan, an ongoing research project) and wants the AI’s suggestions to stay anchored to the same evolving document instead of being re-explained in every new chat.
For technical teams, the more interesting opportunity sits one layer up: OpenAI also opened tools at DevDay for developers to build applications that use agents and operate inside ChatGPT itself. That means the real builder opportunity isn’t necessarily using Spaces as an end user, it’s building the applications and workflows that plug into this collaborative layer, the same way an ecosystem of apps grew up around Slack or Microsoft Teams once those platforms opened to developers.
For anyone evaluating cost, it’s worth noting that OpenAI paired these collaboration features with a cheaper, more efficient model aimed at handling demanding work without consuming heavy usage allowances as fast as the top-tier model does. That matters practically: collaborative, document-heavy workflows generate a lot of back-and-forth, and running that volume on an expensive frontier model would make daily use costly. A more efficient option sitting underneath Spaces and Pages is part of what makes sustained team use financially realistic.
Frequently Asked Questions
What’s the difference between a Space and a Page?
A Space is the shared workspace where people and AI agents collaborate, while a Page is an individual editable document that lives inside that Space. Think of the Space as the shared room and the Page as the document on the table that everyone, including the AI, can edit.
Do Spaces and Pages require the Dots agent to work?
No, but they’re designed to complement each other. Spaces and Pages provide the shared, persistent document layer, while Dots is the proactive agent that can act on that content over time, such as catching outdated information or drafting revisions when something upstream changes.
Is this only for enterprise teams, or can individuals use it?
Other agents start typing. Remy starts asking.
Scoping, trade-offs, edge cases — the real work. Before a line of code.
The collaborative framing targets teams, but the underlying Pages concept (live, editable AI-assisted documents) is useful for individual work too. The bigger enterprise implication is about where teams choose to house ongoing collaborative work, not a restriction on who can use the feature.
Does OpenAI’s agent see everything in my ChatGPT history?
No. Based on how Dots is built, it draws only on memories you’ve explicitly enabled through ChatGPT’s memory settings plus any connected applications you authorize, not your full chat history by default. That boundary is meant to give users control over what context an agent can act on.
How does this compare to Microsoft 365 Copilot in practice?
Microsoft Copilot embeds AI into existing Office documents and apps that enterprises already use. OpenAI’s approach moves the document itself into ChatGPT via Pages, betting that teams will eventually originate work there rather than importing AI into tools built outside of it. Both aim at the same collaborative work surface from opposite directions.
