Is Manus AI the Next Big Thing in Autonomous Agents or Overhyped

The AI world is abuzz with excitement as Manus AI, a general AI agent developed by Wuhan-based startup Butterfly Effect, takes center stage. Unlike traditional chatbots, Manus AI is designed to function autonomously across multiple tasks, integrating multiple AI models like Claude 3.5 Sonnet and Alibaba’s Qwen. But does this mean we’ve finally entered the era of fully autonomous AI agents?

A Bold Step in AI Evolution

Manus AI claims to be the world’s first general AI agent, setting it apart from single-model chatbots like ChatGPT or DeepSeek. Its ability to leverage multiple AI models simultaneously allows it to analyze tasks, navigate the web, and execute complex commands. The concept of an AI that can act as a digital assistant capable of making independent decisions is a game-changer.

Since its launch last week, Manus AI has taken the internet by storm, garnering praise from influential figures in the tech industry, including Twitter co-founder Jack Dorsey and Hugging Face’s Victor Mustar. Some have even compared it to DeepSeek, an AI model that previously shocked the industry with its capabilities.

However, Manus is not yet widely available. Less than 1% of those on the waitlist have received access, despite its Discord community already amassing over 186,000 eager members. MIT Technology Review secured access to the system and conducted a test run—offering a glimpse into both its potential and its shortcomings.

Putting Manus AI to the Test

When tested on three different tasks—identifying top tech journalists, finding real estate listings in New York City, and compiling a list of nominees for MIT’s Innovators Under 35—Manus showcased impressive adaptability but also some frustrating limitations.

Journalist Research Task: The AI initially delivered an incomplete list, failing to provide key details. When prompted, it admitted to “cutting corners” due to time constraints but improved its output significantly upon further instruction. This ability to refine its responses based on feedback is a promising trait for users seeking a collaborative AI assistant.

Real Estate Search: Manus was able to process a complex set of property criteria and returned a well-structured list. However, it struggled with subjective factors, such as evaluating “spacious kitchens” and “good neighborhoods.” Despite this, the AI adapted after multiple refinements, ultimately producing results that felt curated and useful.

Nominee Compilation for Innovators Under 35: This proved to be the most challenging task. Manus AI spent over three hours scouring the web, yet it failed to produce a fully comprehensive list. The AI struggled with sourcing paywalled information, making its research process incomplete.

Is Manus AI Ready for Prime Time?

While Manus AI demonstrates remarkable capabilities, it is not without flaws. The system frequently encounters stability issues, struggles with restricted content, and occasionally crashes under heavy loads. Additionally, its self-admitted tendency to “rush” through tasks raises concerns about reliability.

Despite these limitations, Manus AI represents a major step toward AI autonomy. If its creators can refine its decision-making, enhance stability, and optimize data sourcing, it could become an indispensable tool for businesses and professionals alike.

For now, Manus AI remains a work in progress—an exciting glimpse into the future but not yet the revolutionary AI agent it aims to be.