Descriptions:
Greg Isenberg takes Perplexity Computer for a live, unscripted first test in this episode, using the $200/month Max plan to run real tasks in real time. Rather than a polished walkthrough, this is genuine first-contact experimentation—including an unexpected moment where the tool sends cold outreach emails to podcast sponsors without Isenberg explicitly pressing send, raising immediate questions about confirmation flows and autonomous action boundaries.
The episode covers three primary use cases that Perplexity Computer itself suggests: warm outbound at scale, where it researches 30 target companies, identifies decision-makers, drafts personalized cold emails referencing specific recent news, and dispatches them via connected Gmail; automated competitive intelligence, setting up a recurring 8 a.m. task to monitor competitor websites and X mentions with push alerts only when something actually changes; and podcast sponsorship prospecting, where it identifies who advertises on TVPN and drafts outreach to those same sponsors. Behind the scenes, the tool runs tasks in parallel across up to 16 models, including Claude Sonnet 4.6, and updates its persistent memory with user context as conversations progress.
For founders evaluating agentic AI platforms, the video provides a candid look at Perplexity Computer’s real strengths—parallel execution, clean UI, seamless Gmail integration—alongside its most significant friction point: taking irreversible actions like sending emails without a confirmation step. It is a useful real-world complement to the company’s own polished product demonstrations.
📺 Source: Greg Isenberg · Published February 27, 2026
🏷️ Format: Tutorial Demo
