The context tax
Every productivity suite contains the same hidden chore: explain the project again. The decision is in chat, the specification is in a document and the deadline is in a tracker. An assistant attached to only one surface sees a fragment.
AI-native workspace builders argue that the agent should sit above the fragments—or that the fragments should disappear into one shared data model. That promise explains why Core’s compact launch message resonated so strongly.
A product problem, not only a model problem
Centralized context can improve answers, but it can also centralize noise. Teams need durable distinctions between a proposal, a decision, a task and an informal message. Without that structure, an agent gains more tokens and less clarity.
The best workspace will not be the one that stores everything. It will be the one that understands which information is authoritative, who can change it and when it expires.
How to evaluate the new category
Do not begin with a company-wide migration. Choose one recurring workflow, such as turning a customer call into a prioritized issue and a drafted follow-up. Track how many manual handoffs disappear and how often the agent uses stale context.
Open-source options add a useful question: can the team inspect permissions, data flow and agent actions? In a workspace that may contain the company’s memory, observability is part of productivity.
Sources & further reading
Social-media activity is treated as a signal of attention, not proof. Product claims are attributed to the linked publisher or announcement.