Usable AI agents are finally here. From OpenAI’s DeepResearch to xAI’s DeepSearch, we’re seeing the first real push toward autonomous tools that can plan, execute, and complete tasks like research, outreach, and coding with minimal human input. Recently, a new entrant joined the competition — Manus, a general-purpose AI agent that’s getting global attention. Some are calling it the “most sophisticated computer using AI,” while others write it off as just another “wrapper.” So what sets Manus apart under the hood? What does its launch say about the broader debate around GPT “wrappers”? YC’s @garrytan dives into Manus’ “chain-of-thought injection” technique that enables agents to actively update plans, what works particularly well with its UI, and how founders can create enduring value with application-level products.
