OpenAI’s fourth DevDay brought 22 product and feature launches. Two of them, ChatGPT Space and Astra Ultrafast, change different parts of using ChatGPT: where people work with AI, and how quickly AI produces text. Space brings shared files and editing tools into ChatGPT so people and agents can work together. Astra Ultrafast is a speed tier available on a new $500-per-month plan. The first raises the question of whether teams can keep their work in one place; the second, whether faster output is worth faster token consumption. Tokens are the small units used to measure generated text.
Space puts people and agents in the same workspace
ChatGPT Space combines cloud file storage, similar to Google Drive, with editors for documents, presentation slides and spreadsheets inside ChatGPT. Instead of treating a chat response as something to copy into another app, a user can work on the file in the same environment where an AI agent is available. People and agents can collaborate concurrently, making the shared file, not a separate conversation, the center of the task.
▲ Human and agent collaboration in Space
A document comment offers a direct way to involve an agent. A user can tag a dot in the comment, prompting that agent to carry out edits. OpenAI’s dots are persistent agents in ChatGPT: each one sits in a dedicated tab at the top of the desktop and mobile apps and stays available across interactions, rather than requiring a new chat thread for each task. In Space, that persistence matters because the agent can join work that people are already doing on a document.
The approach could reduce the handoff between asking AI for help and applying its output. A team could keep its files, comments and editing activity within ChatGPT rather than exporting work to Notion or Google Workspace. That is the intended advantage, not yet proof that Space can replace an established workflow. It remains uncertain whether its editors have enough advanced features and polish to displace tools such as Google Docs, Google Sheets, Keynote or Notion; that will take weeks of real use to judge. The useful early test is whether the file types and comment-based agent collaboration cover the work a team actually needs to do.
Early dots still have rough edges
A dot can use its own virtual machine as well as the user’s computer, and it acts proactively, for example by sending updates or confirming that it replied to a meeting request. In early-access testing before the keynote, however, dots showed noticeable bugs, including confusion between the in-app browser and the cloud computer and failed logins on some websites. Before relying on dots for critical work, waiting one to two weeks gives OpenAI time to ship stability fixes.
Astra Ultrafast trades waiting time for token use
Astra Ultrafast addresses a different friction: the wait while a model generates output. It is a speed tier that makes generated text arrive much faster, which can make an interaction feel more immediate. That speed does not guarantee how long a particular task will take or how good the answer will be.
▲ Fast output and token consumption
Access comes with a new $500-per-month plan aimed at users who need substantially more headroom than the existing $200-per-month Pro plan. The plan price and the speed are separate considerations. A faster stream of output can also use a token allocation exceptionally quickly, so speed alone does not say how long a user’s available tokens will last.
For someone considering the higher-priced plan, the practical test is to compare the value of the faster response with the tokens consumed while doing real work. Monitor usage during a limited trial rather than treating top speed as the whole experience. Space and Astra Ultrafast solve different problems: one changes how work is shared, while the other changes how quickly output arrives. Because both are built into ChatGPT, they may give OpenAI a structural edge over standalone tools, and they fit a broader push to make ChatGPT a hub for everyday work and for developers’ products.
What to test before changing a workflow
Start with the bottleneck that matters. If collaboration is the issue, try Space with the document, slide or spreadsheet work your team already does. Check whether people can make the edits they need and whether tagging a dot in a comment helps without forcing work back into another tool. Keep existing tools in place until Space’s editing depth meets those needs, and hold off on handing critical work to dots until the early bugs are fixed.
If waiting for generated output is the issue, assess Astra Ultrafast against both its $500 monthly access price and your token consumption. Faster output may make interaction more responsive, but the quicker use of an allocation is part of that experience. Neither a shared workspace nor a faster tier settles the decision on its own; the right test is whether it improves the work you need to complete.