Artificial intelligence is evolving at a breakneck pace, and in just the past few days, we’ve seen numerous developments fly under the radar—except one. Manus AI has exploded in popularity, amassing millions on its waitlist and fueling speculation that it’s a significant leap forward, possibly even a glimpse into artificial general intelligence (AGI). But is it really as revolutionary as the hype suggests? Let’s break it down.
The Perfect AI Hype Campaign
Before diving into the technical details, let’s analyze what makes a perfect AI hype campaign. Manus AI’s rollout follows a textbook example:
- Marketed as a paradigm shift in AI, hinting at AGI capabilities.
- Access restricted via a waitlist, with early access granted to those likely to generate buzz.
- Invite codes become scarce, creating a black market where they’re resold for thousands of dollars.
- The core technology is kept vague initially (later revealed to be based on Claude 3.7 Sonet).
- Benchmarks are selectively shared to highlight strengths while avoiding weaknesses.
This approach masterfully stokes public curiosity while maintaining an aura of exclusivity.
What is Manus AI, and How Does It Work?
At its core, Manus AI functions as a sophisticated AI agent combining multiple tools and models. Think of it as an enhanced version of OpenAI’s Operator and Deep Research systems:
- Operator: An AI agent that can take actions, such as booking restaurants and automating workflows.
- Deep Research: An AI system that scours the web, analyzes sources, and provides detailed insights.
- Claude Projects: A framework for interactive previews of AI-assisted work.
By merging these capabilities, Manus AI creates an experience where users can automate complex research tasks, interact with dynamic outputs, and even execute actions. It’s an ambitious concept, but does it truly deliver?
Testing Manus AI Against Its Rivals
To gauge its effectiveness, I compared Manus AI with similar tools:
- OpenAI’s Deep Research
- Google’s Deep Research (Gemini Advanced)
- Grok 3 Deep Search
One test involved asking these AIs to analyze an image and list the founders of each company featured in it. Here’s how they performed:
- Gemini Advanced Deep Research: The fastest, but with limited functionality—couldn’t process files.
- Grok 3 Deep Search: Took 2.5 minutes, analyzed 344 sources, but missed many company founders.
- Manus AI & OpenAI Deep Research: Both took around 15 minutes, with OpenAI’s model producing more reliable results.
In another test, I tasked each model with researching and comparing one another. Manus AI provided a comprehensive table but struggled with accuracy, omitting key details about its own benchmark scores and pricing.
The Cost Factor: Is Manus AI Sustainable?
One of the biggest red flags is the cost. Estimates suggest that each Manus AI query costs around $2, making it prohibitively expensive for everyday users. If the service is ultimately priced at around $200 per month, as some speculate, it will be out of reach for most consumers, limiting its widespread adoption.
Hype vs. Reality
Manus AI fits into a broader pattern emerging in AI for 2025:
- AI models are becoming more expensive.
- Performance remains inconsistent.
- Despite impressive moments, many AI tools struggle with reliability.
- Marketing and hype play a bigger role than ever.
The truth is, while Manus AI is an innovative combination of existing tools, it does not represent a breakthrough akin to DeepSeek or GPT-5. It’s a compelling experiment in AI-assisted automation, but its limitations, cost, and accuracy issues suggest that the hype may be overstated.
What’s Next for AI?
As AI continues to advance, expect more companies to use similar marketing strategies—exclusive waitlists, influencer campaigns, and selective benchmarks. The success of Manus AI’s hype campaign ensures that this trend will persist.
For now, if Manus AI does end up costing several hundred dollars a month, I’d personally give it a miss. However, if you’re interested in exploring AI in a more hands-on way, consider looking into red-teaming challenges like the Grace One competition, which helps improve AI reliability while offering substantial cash prizes.
Let’s keep an eye on what’s next—but let’s also stay skeptical of the hype.