Perplexity Projects adds durable shared workspaces to Computer
Perplexity's 12 October research update explains persistent project files and concurrent work. Internal retrieval benchmarks do not establish commercial retention or gross margins.
Latest evidence review — 16 October 2026
Perplexity's 12 October research publication describes Projects as shared workspaces for people and agents undertaking long-running Computer tasks. Files persist across sessions, contributors work on a common history, and conflicting changes are coordinated through version control. This adds a durable collaboration layer to the recurring automations covered below.
The paper also reports an internal comparison between file exploration and semantic retrieval on financial questions. That is a bounded engineering benchmark, not a measurement of all Computer tasks, customer productivity or commercial conversion. Its results should not be generalized into a revenue or margin forecast.
Our assessment is that persistent workspaces could increase switching costs if organizations accumulate useful context and repeatable processes. They also raise questions about storage, permissions, recovery, operating costs and responsibility for conflicting edits. The investment case requires evidence that customers continue paying for the workflow after initial experimentation. Product architecture provides a reason to examine those questions, without answering them. The source expands the description of the business but supplies no confirmed IPO calendar, public share price or completed offering.
Source: Perplexity shared workspaces — 12 October 2026; Perplexity Research dated index
An ongoing product relationship
Perplexity’s 5 October 2026 update added automations to Computer for scheduled work and events in connected applications. The change creates a different use pattern from a single search: a customer can return to an ongoing task rather than begin each request from scratch.
Our interpretation is that recurring work could strengthen the relationship with the product if it reliably saves time. That opportunity comes with a more demanding operating test. The service must handle changing inputs, maintain permissions and recognize when a person should review an action. A product launch does not establish how consistently it meets that test.
What investors would need to learn
The IPO watchlist still reflects the chief executive’s reported ambition for 2028. No confirmed offering calendar, ticker or price was identified by 7 October.
The automation release makes retention and cost disclosure more important to the eventual investment case. Useful questions include how many customers maintain recurring workflows, how often tasks complete successfully and how the cost of longer work compares with subscription revenue. Product activity should also be separated from paid adoption.
A public prospectus could provide that financial context. Until then, October’s release is a concrete expansion of the business model under consideration, while the timing and valuation of a potential listing remain open.