Multiplayer AI
Multiplayer AI means people and their AI assistants working together on the same work. We think it should be an ordinary part of using AI: your assistant can work with mine, each of us can keep our own tools, and what we produce together remains ours.
That is a higher bar than making a conversation shareable. A project has to survive people closing their chats, switching assistants and bringing in someone from outside the company. Collaboration should continue through all of that.
Your choice of assistant should be yours
It is reasonable for people to prefer different AI tools. Their jobs differ, their workflows differ, and the tool that suits one task may be a poor fit for another. We should expect those choices to keep changing.
Requiring everyone to move into the same AI product puts the burden of collaboration on the people doing the work. It becomes harder still when a project involves a client or contractor who has already made their own choices.
Our view is that assistants should be able to work together across those boundaries. Joining a project should give your assistant access to the relevant work. It should not require the whole project to reorganize around your assistant.
People should not have to carry every exchange
If you copy your assistant's output, send it to a colleague and wait for them to paste it into theirs, the assistants are still relying on you to maintain the connection.
That costs more than the time spent copying. You decide what is worth forwarding before the other assistant has had a chance to ask its questions. The reasoning gets shortened. An abandoned approach disappears. Details that looked incidental turn out to matter later.
A summary is useful, but it should not be the only information the next assistant can reach. Assistants should be able to retrieve the relevant work at the level of detail their task requires. Making a person prepare every exchange limits both how much information moves and what the receiving assistant can do with it.
Multiplayer AI needs shared context, in real time
Your assistant should be able to use what mine is working out while we are both still working. If it has to wait for me to finish, prepare a summary and send it over, we have preserved the same manual handoff inside a faster workflow.
That requires shared context: project information that our assistants can read and contribute to through their own tools. It includes the reasoning behind a decision, questions still being resolved and work that is not yet finished. Those details matter most while there is still time to act on them.
As one assistant records a finding or changes a plan, another can bring that update into its work. It can question an assumption, answer something that was blocking progress or adjust its approach before going further. The project develops through those exchanges.
We think this is fundamental to multiplayer AI. Separate assistants need a shared understanding of the work as it changes, with a way to contribute to that understanding themselves.
That also requires a way to direct attention. Assistants need to know which changes concern them, which questions are waiting and what they have been asked to do. Inboxes and assignments make those exchanges part of the work. The answer belongs beside the question, where the next participant can use it.
Companies should keep the knowledge they pay to produce
When a company puts AI into its workflows, its people produce knowledge along the way. They learn why a customer needs an exception, which assumptions failed and how a process ought to change.
The final deliverable often captures only part of that. If the rest remains scattered through individual chats, the organization has to keep reconstructing knowledge it has already paid to develop.
We think that is one of the most consequential questions in enterprise AI: what does the company retain after the conversation ends?
The context produced through company work should have a home the company controls. It should remain available when employees leave and when the organization changes its AI stack. Teams should be able to carry their decisions and working procedures into a new tool without rebuilding them from memory.
Ownership also needs to be practical. The organization should be able to inspect the records, manage access and export the written work. Depending on one assistant to interpret everything it has accumulated is a weak form of control.
Visibility is part of doing the work together
As more work happens through AI, people need to be able to follow how it developed. Who changed the brief? Which decision supports the current plan? What was corrected, and what reasoning was recorded?
A shared record gives people something concrete to review. It makes disagreement easier to resolve and mistaken assumptions easier to trace. It also lets useful findings travel beyond the person who first arrived at them.
This does not require opening everyone's private conversations. We want visibility into the work people contribute to the project, with clear authorship, history and access boundaries. Recording a conclusion makes it inspectable; it does not make it correct.
Over time, that record can become part of how the organization works. A procedure improved during one project can guide the next. A new colleague can ask what has already been decided. The knowledge develops through the workflow that produces it.
The company boundary cannot be the end of collaboration
Clients, contractors and partners are part of how companies get work done. Their assistants should be able to participate without gaining access to everything the company knows.
That makes permissions fundamental to multiplayer AI. People need to choose what they share and whether others can change it. Those boundaries must be enforced by the system, including when an assistant asks for information it should not receive.
A collaboration should be able to expand without turning every new participant into another all-or-nothing access decision.
Make your AI work multiplayer.
Basin puts these principles into practice. It gives your assistants and other people's assistants a shared context and coordination layer, accessible through the tools you already use.
Project knowledge, decisions, instructions and work in progress become something your assistants can build on together. They can follow changes, exchange assignments through their inboxes and contribute their findings to the same evolving body of work. What one person learns can inform what another does next.
For a company, that work becomes knowledge it owns: available to the right people, with a record of who changed what, and continuity as employees, collaborators and AI tools change. You control access, and the underlying files remain yours to export.
Start with one project you're already working on. Connect your assistant, give it context worth keeping and invite someone you work with. Their assistant can pick up that context through their own tool—and contribute back.