An Example of Where Manus Excels

I've written before on Manus.im and why I think it's a great tool. Here, I want to provide an explicit example.

I had a bunch of original content that I wanted to turn into a long-form blog post (going out tomorrow) that shares how I white labeled my system in a week, front to back. The content was about 8,000 words long, mixed with code and text.

I dumped it into ChatGPT and asked for a cleanup, focusing on structure, organization, etc while leaving nothing out. I set the context:

I have a blog called "Technically Speaking" where I blog about my experience as a technical leader and builder. Please see my posts here: https://heavychain.org/blog. It's a mixture of technology, math, musings, and leadership.

I need you to help me author a gorgeous post about this white labeling experience I just went through. We should be (1) correct, (2) thorough, (3) detailed, (4) and explanatory. The value here is in helping people learn how to design systems in this case a white label system.

I can provide architecture diagrams and screenshots later -- feel free to call those out in the text.

I want you to review my rough draft above. It is very long. If you need to break it down into sections.

Once reviewed and understood, please author the complete article by cleaning up what I shared. Again, we're after a well written, long-form post that takes its time in ushering the reader through a fantastic engineering challenge and solution.        

ChatGPT's Response

ChatGPT, using Canvas only returned a mere portion of my original content. It essentially provided an introduction and then a short summary. It was 1,008 words long. It supplied 4 code blocks. I took my prose and converted most of it to bullets, despite my asking the opposite. I needed the long-form content preserved and cleaned up, not elided.

Also, ChatGPT changed my writing style from -- my own brand voice -- to standard LLM output. I got a lot of "X isn't just this—it's that!" and similar styled text. It screamed, I just copy pasted from an LLM.

Manus.im's Response

I gave the same prompt and context to Manus AI . Manus knocked this task out of the park. It adopted my tone of voice, hewing very closely to it. It kept much of what I wrote verbatim, too, mostly reorganizing and cleaning up as asked. The long-form ask was adhered to, even a little too well. I ended up cutting it down quite a bit.

Here was Manus's summary of it's work for me:

I've completed your comprehensive blog post about building a white label system! The article is over 11,000 words and follows the narrative, technical style of your "Technically Speaking" blog.

Here's what I've created:

Structure & Content:

• Introduction with the NewRoots.ai context and white labeling challenge
• Requirements definition and architectural constraints
• Frontend architecture with domain-based tenant detection
• Backend architecture with tenant-aware services and data isolation
• Report generation system with dynamic PDF branding
• Email system using MJML templates with tenant variables
• Comprehensive testing and validation approach
• Lessons learned and architectural reflections
• Conclusion tying together business and technical impact

Writing Style:

• First-person narrative matching your blog's conversational tone
• Technical depth with practical code examples
• Honest about challenges and trade-offs made under pressure
• Story-driven structure with clear lessons learned
• Mix of technical implementation and business context

Technical Coverage:

• React Context and dynamic theming with DaisyUI
• FastAPI middleware and dependency injection
• Database design for multi-tenancy
• Pandoc/LaTeX templating with Jinja2
• MJML email templates with tenant branding
• End-to-end testing strategies
• Performance and security considerations

The article captures the engineering challenge and solution while maintaining the engaging, educational tone that makes your blog posts so valuable for other technical leaders.

The only thing Manus didn't do well was redundant expatiation. It was so focused on length that it expanded one of the sections with redundant paragraphs. I cut that out.

So, as a quick summary. ChatGPT didn't even get the task correct. Manus, however, nailed it. It handled the content, language, code, length, everything. And, when it was done cleaning things up, I was feeling lazy at this point and I asked it:

Manus, literally browsed to the right location, fetch the right screenshots and included them in the right place in the doc.

Manus AI is amazing and has its place the the larger AI ecosystem.

Still a fan.