One with my original thought, tone, and verified data in a formal structure:
75 pages, structured chapters, index, acknowledgements, bio, and cover.
It’s not an easy nor simple task for AI
I put the LLMs to test, and here are the scorecards:
Perplexity
Felt like a smarter research assistant.
It pulled quotes, case studies, and recent articles to back up my ideas. Great for reference checking and grounding data.
But it never understood the through line. No structure, no tone consistency, and no sense of creative intent. It was like having an analyst who brings you every report, but doesn't help you tell the story.
Use case strength: Fast factual recall, zero narrative cohesion
Score: 7/10
ChatGPT
This one felt like a well-read co-writer.
It gave me thoughtful phrasing, strong section intros, and even helped refine arguments.
For short essays or frameworks, it’s excellent. Quick, articulate, and stylistically flexible.
But for a full 75-page book, it lost memory fast. By Chapter 3, tone drifted. By Chapter 10, logic unraveled.
Use case strength: Short-form content, structured ideation, clean prose.
Score: 7/10
Claude
Organized, logical, and very “editorial.”
It grouped chapters, tightened paragraphs, and added empathy to my tone something I appreciate as a storyteller.
But it hesitated constantly, rewriting instead of shipping.
Claude felt like a cautious editor with great judgment but no deadlines. Every time I tried to push for progress, it wanted to revise.
Use case strength: Editing, refinement, summarization.
Score: 6.5/10
Manus AI
This one was different. It built the book.
From chapter outline to layout, index, and acknowledgements, all in a single workflow. It executed without extensive multi prompt steps.
It assigned sub-agents for structure, tone, design, and formatting
I could change or add instructions while it was still working, dynamic prompt editing
I could redirect tone, swap examples, or reframe an entire section in real time, which is helpful given the complexity of creating a book.
Use case strength: Multi-step execution, complex creative orchestration, live iteration.
Score: 8.5/10
As a GTM leader, I’m looking for agentic AIs that act like seasoned executive assistants who can collaborate and deliver actions and outcomes.
In that world, we can all accomplish more, with quality.
📸 Spoke about my experiences at the Manus AI “Vibe and Ship IRL” event during SF Tech Week.
