Perplexity Moves Deep Research Into Computer, Routing Research Subtasks Across 20+ Frontier Models For Reports, Decks, And Dashboards

- Perplexity just shoved Deep Research into Computer, routing subtasks across 20+ frontier models. It’s not a moonshot; it’s an agentic search pipeline that writes code to fetch data, then compiles reports, decks, and dashboards. BrowseComp accuracy jumped from 40.7% to 83.8%, which is impressive if you trust first-party benchmarks. It reads your PDFs, cross-references Statista, and cites everything. For crypto analysts, this means less time hunting for alpha and more time verifying if the "alpha" is actually just hallucinated noise. The Agent API is open for devs, so expect a flood of automated research bots soon. Don't let the shiny deliverables fool you: cited doesn't mean correct. Verify the data before you stake your hash manifest on it.