This video provides a first-look review and test of Comet, a new "AI-first" browser from Perplexity. Built as a fork of Google Chrome, Comet offers a fast, seamless setup by importing all existing user settings, extensions, and authentications. The browser's core innovation is its integrated AI agent, which operates locally rather than in the cloud. This local execution is highlighted as a key advantage, as it allows the agent to interact with websites where the user is already logged in and to take over tasks in progress, eliminating major points of friction common to other web automation tools. The review demonstrates Comet's capabilities through a series of tests, including managing LinkedIn requests, checking real-time product inventory, and posting to social media, showcasing both its successes and current limitations. The video concludes by positioning Comet as a glimpse into the future of web interaction, where users delegate tasks to AI agents for greater efficiency and to navigate an increasingly complex online environment.
A First Look at Perplexity's AI Browser, Comet: The Future of Web Browsing?
0:00:03 The Future of Browsing: The video introduces Comet, Perplexity's new AI browser, proposing a future where users task AI agents to browse the web for them. This new interaction model is referred to as "vibe browsing."
0:00:52 Seamless Chrome Integration: Comet is a fork of Google Chrome, ensuring a frictionless setup. It automatically imports all user settings, bookmarks, extensions, and, crucially, authenticated sessions from Chrome. The reviewer notes it feels surprisingly fast.
0:03:00 The Advantage of a Local Agent: A key differentiator for Comet is that its AI agent runs locally on your machine. This solves a major pain point of cloud-based agents by using your existing logged-in sessions, so it doesn't need to be separately authenticated. It also allows the agent to have the context of what you're currently doing on a webpage.
0:04:43 Core AI Features: The browser deeply integrates Perplexity, making it the default for search and new tabs. The main interface is an assistant button that can be invoked at any time. You can @mention open tabs to have the AI perform tasks on that specific page, such as summarizing a YouTube video.
0:07:24 Task Automation in Action: The reviewer tests Comet on several real-world tasks with mixed results:
LinkedIn: It successfully automated the process of accepting all pending connection requests.
YouTube: It failed to post a comment reply, citing "technical restrictions" from the website.
Twitter (X): It successfully posted a quote-tweet after getting user confirmation but added an unrequested "created with comet assistant" signature to the post.
0:09:17 Explicit Instructions are Key: When first asked to find a Nintendo Switch 2, the agent acted like a search engine and just returned links. However, with a more explicit command to "Go to each of those websites and check it for me," it successfully performed the automated inventory check.
0:11:54 Live Browser Control: The agent can take direct control of the active browser window to perform tasks, such as navigating to a link on the page. While the process can be slightly clumsy, it demonstrates the ability to execute commands in the user's immediate view.
0:13:31 A New Web Paradigm: The creator concludes that Comet is a strong signal of a future where users interact less with websites directly and more with AI agents. This paradigm shift could enable users to perform many tasks in parallel and use AI to filter through the growing "slop" of AI-generated content on the internet.