Manus AI Review: Like a Robot, but Online
Hype or Real -- You decide
Quotes from the Story:
Is Manus truly the paradigm shift it claims to be, or just another overhyped AI product?
These materials provide everything I need to systematically reach out to communities that might be interested in "Reader/Rocketship" while tracking my progress and maintaining a professional approach. Amazing.
It offered to produce a website, which I thought would be a copy; what it turned out to be was a website which provided the documentation: The Marginalian Analysis, designed to adopt some of the style of the original.
At some point soon, we will have enough confidence to give our passwords to AI agents to access our email accounts so that agents can work for us without us having to touch a keyboard or tap on a screen.
Make our reading count.
Call to Action Throughout:
Subscribe to get updates directly from Substack.
Share forward through Sqale. We can avoid the machinery of marketing.
Join our fortnightly Invitational Session (via Eventbrite).
How Manus AI Shapes Up
I've been tracking AI developments closely for years, and when Manus AI burst onto the scene in March 2025, my initial reaction was “generic agent, getting closer to something I can actually use”. The demos looked impressive—promising an AI that could actually execute tasks rather than just chat about them.
Then I got early access today (7th April), and everything changed.
This isn't just another AI chatbot—Manus represents a genuine evolution in what these systems can do. While other AI tools excel at answering questions, Manus takes a significant leap toward actually "doing stuff for you." It's the difference between having a knowledgeable advisor and having a capable assistant.
The experience has been eye-opening, but not without raising important questions. Manus' pricing model brings to the surface inevitable issues around AI accountability and contract disputes that we'll need to address as these agentic solutions become integrated into our daily lives and work. When an AI is empowered to act on your behalf, who bears responsibility when things go wrong?
For context, Manus arrived at an interesting moment in AI development. Anthropic had revealed their Computer Use capabilities in October 2024, taking a developer-first approach. In contrast, Manus AI—built by a team in China—surprised the entire community with their bold consumer-facing strategy from Day 1. Many industry observers (myself included) questioned whether their capabilities could possibly match their promises.
How does it stacks up against the current AI landscape, including Google's Gemini, Anthropic's Claude, Perplexity, and OpenAI's ChatGPT. Is Manus truly the paradigm shift it claims to be, or just another overhyped AI product?
Here are two tasks I asked it to perform, while sharing two links to promote content I think is amazing.
First Task: Find Folks Interested in A New Article
I put this prompt in:
Take a look at this innovative short story:
What kind of groups, sites, forums, boards, communities might be interested in it? Can you contact them on my behalf, or provide detailed info on how to contact them?
These are the materials it provided (check the original Manus thread if you want):
Outreach Tracking Spreadsheet (outreach_tracker.csv) - This spreadsheet contains:
All 18 communities organized by category
Complete contact information for each community
Columns to track outreach status, responses, and follow-up dates
Notes specific to each community to guide your approach
Tailored Messages (tailored_messages.md) - I've created fully customized messages for each community:
All placeholder fields have been replaced with specific information
Each message is tailored to the interests and format of the specific community
Messages are organized by community type for easy reference
Different formats are provided for Reddit posts, forum discussions, emails, and social media
To execute the outreach strategy:
Import the CSV file into your preferred spreadsheet program (Excel, Google Sheets, etc.)
Start with the communities marked as highest priority in the notes
Copy the corresponding tailored message from the messages document
Send the message through the contact method specified in the spreadsheet
Update the spreadsheet with the date contacted and any responses
Follow up as needed based on the strategy in the original outreach document
These materials provide everything I need to systematically reach out to communities that might be interested in "Reader/Rocketship" while tracking my progress and maintaining a professional approach. Amazing.
Currently Manus can not send emails on my behalf, nor post on various communities and groups online, both of which need to be done manually at the moment.
Took less than 10 mins.
Second Task: Analyse a Website
The prompt:
Please review the content of this site:
https://www.themarginalian.org/
I think it is the most brilliant newsletters I have ever read. Can you summarise style and content choice in a comprehensive analysis so that it can be used as a guide for prompts for AI to reproduce posts that have a similar style as well as depth and range of content choice?
It came up with several documents:
A comprehensive analysis document that details the site's structure, writing style, content themes, and article patterns
A practical AI prompting guide with specific instructions, sample prompts, and evaluation criteria
Detailed notes on individual articles that informed the analysis
It offered to produce a website, which I thought would be a copy; what it turned out to be was a website which provided the documentation: The Marginalian Analysis, designed to adopt some of the style of the original. Check it out.
Conclusion: The Beginning of Useful Web-Agents
The speed of development means that by the end of the year (2025), every AI service will be offering generic internet-search-enabled AI services. At time of writing, Claude’s internet-enabled search feature is locked behind the second-tier payment system: you need five accounts each paying $25.
At some point soon, we will have enough confidence to give our passwords to AI agents to access our email accounts so that agents can work for us without us having to touch a keyboard or tap on a screen. Ask it to do anything online for you, and it does.
Like a robot, but online.
Invitation to Read & Review
Thanks for reading! This story conforms to the practices of Fulcrum : Reflexive Reading, Organic Sharing and A-State self-organisation. In the face of growing AI, our organic readership strengthens our social ties over this crucial decade. Share forward through Sqale so we can avoid the machinery of marketing, and find and support writers who can contribute to this narrative. Make our reading count.
How Does Sqale Work?
Money is converted into Credits on Sqale. You can’t use Credits to buy anything. Use Credits to share what you value. We call this Organic Sharing.
Use Sqale to share anything you value whether it is charity work, music, films, games, anything — share what you value with people you value.