Generate long-form text content with Manus and Gemini Deep Research
Here’s something I’ve done with Manus. Task: Here’s something I’ve done with Manus. Manus played a critical role in the recent work I’ve done with long form text content generation. I developed a specific type of prompt engineering i call “fractal iteration” it’s a specific method of hierarchical decomposition which is a type of top down engineering.Using my initial research and testing, here is a long form prompting guide I developed as a resource. It’s valuable to read, but equally valuable as a tool to create a prompt engineering LLM. https://towerio.info/uncategorized/a-guide-to-crafting-structured-deep-long-form-content/ This guide can produce really substantial work, including the guide itself, but it actually gets better.When a style guide and planning structure is used, it becomes incredibly powerful. Here is a holistic analysis of a 300+ page nonfiction book I produced with my technique, as well as half of the first chapter. I used Gemini Pro 2.5 Deep Research and Manus. Please note the component about depth and emotion. https://pastebin.com/raw/47ifQUFx And I’m still going to one up that. The same methods and pep materials were able to transfer the style, depth, and voice to another work while maintaining consistency, as the appendix was produced days later but maintains cohesion.I was also able to transfer the style, voice, depth, and emotion to an equally significant collection of 100 short stories over 225,000 words, again using Gemini and Manus. https://mvcc.towerio.info/ And here is an analysis of those stories: https://pastebin.com/raw/kXhZVRAB Manus played a significant role in developing this content. It can be easy to say, “oh well it’s just because of Manus” and I thought so maybe as well, but detailed process analysis definitely indicates it’s the methodology and collaboration. I kept extensive notes through this process. Huge shoutout to Outskill, Google, Wispr Flow (my hands don't work right to type), aiToggler and Manus for supporting this work. I’m a profoundly disabled brain tumor survivor who works with AI and automation to develop assistive technology. I have extremely limited resources - I was homeless just two years ago. There is absolutely still so much to explore with this and I'm really looking forward to it!; tools: Manus, Gemini. Human role: The original post does not say how a person stayed involved.; limitations: Prerequisites include access to Manus and Gemini Pro 2.5 Deep Research. The source is a self-reported Reddit post with no external verification of output quality or consistency. Integration uncertainty exists regarding how 'fractal iteration' scales without the author's specific prompt engineering expertise. Human approval is required for all content before publication..
Tools
Published on Agent Work Atlas:
Overview
Here’s something I’ve done with Manus.
Workflow
Here’s something I’ve done with Manus. Manus played a critical role in the recent work I’ve done with long form text content generation. I developed a specific type of prompt engineering i call “fractal iteration” it’s a specific method of hierarchical decomposition which is a type of top down engineering.Using my initial research and testing, here is a long form prompting guide I developed as a resource. It’s valuable to read, but equally valuable as a tool to create a prompt engineering LLM. https://towerio.info/uncategorized/a-guide-to-crafting-structured-deep-long-form-content/ This guide can produce really substantial work, including the guide itself, but it actually gets better.When a style guide and planning structure is used, it becomes incredibly powerful. Here is a holistic analysis of a 300+ page nonfiction book I produced with my technique, as well as half of the first chapter. I used Gemini Pro 2.5 Deep Research and Manus. Please note the component about depth and emotion. https://pastebin.com/raw/47ifQUFx And I’m still going to one up that. The same methods and pep materials were able to transfer the style, depth, and voice to another work while maintaining consistency, as the appendix was produced days later but maintains cohesion.I was also able to transfer the style, voice, depth, and emotion to an equally significant collection of 100 short stories over 225,000 words, again using Gemini and Manus. https://mvcc.towerio.info/ And here is an analysis of those stories: https://pastebin.com/raw/kXhZVRAB Manus played a significant role in developing this content. It can be easy to say, “oh well it’s just because of Manus” and I thought so maybe as well, but detailed process analysis definitely indicates it’s the methodology and collaboration. I kept extensive notes through this process. Huge shoutout to Outskill, Google, Wispr Flow (my hands don't work right to type), aiToggler and Manus for supporting this work. I’m a profoundly disabled brain tumor survivor who works with AI and automation to develop assistive technology. I have extremely limited resources - I was homeless just two years ago. There is absolutely still so much to explore with this and I'm really looking forward to it!
Reported result
The author reports producing a 300+ page nonfiction book and a 225,000-word collection of 100 short stories. They claim the method maintained style, depth, and voice consistency across parts generated days apart. No independent metrics or measured outcomes are provided; success is attributed to the collaboration between the specific methodology and the AI tools.
Human role
The original post does not say how a person stayed involved.
Limitations
Prerequisites include access to Manus and Gemini Pro 2.5 Deep Research. The source is a self-reported Reddit post with no external verification of output quality or consistency. Integration uncertainty exists regarding how 'fractal iteration' scales without the author's specific prompt engineering expertise. Human approval is required for all content before publication.
Try it
I want to test a long-form content generation workflow using Manus and Gemini Pro 2.5 Deep Research. My goal is to produce a structured [replaceable: e.g., 50-page guide or 10 short stories] on [replaceable: topic]. First, help me define a 'fractal iteration' prompt structure: break the topic into hierarchical sections, then create a style guide and planning outline. Use this outline to generate a small sample section (approx. 1,000 words) via Gemini Deep Research for factual grounding, then refine the voice and depth using Manus. Do not generate the full content yet. Present the sample and the prompt structure for my review. I will confirm if the tone and depth meet my standards before proceeding to the next batch. Ensure all outputs are saved to [replaceable: local folder or cloud drive]. If the sample lacks emotional depth or structural coherence, ask me for specific adjustments to the style guide before retrying.
Source reports are not independently verified. Publication date is the catalog publication date, not the original article date.