Creating an Outline and Authoring Chapters - Self-Publishing 4D PKM in 6 Weeks - VLOG Episode 4

A self-publishing author is turning a 15-chapter, ~9,000-character book outline into a full draft by running an iterative, tool-assisted workflow—using Gemini, NotebookLM, and a paid ChatGPT setup—to repeatedly refine structure, then generate chapters quickly enough to target a March 15 launch.

The project’s backbone is a four-part arc that starts with explaining why the reader is “in the situation” they’re in, backed by cognitive science and learning history, before moving into a solution. The outline is already assembled and deeply detailed—so detailed that it can’t be treated as a simple plan. Instead, it becomes a living knowledge base that gets condensed, reworked, and fed back into the writing system as the author cycles through feedback and reference material.

Early attempts at writing through a visual workflow proved too slow and too hard to reuse. Card-based thinking helped generate ideas, but it didn’t “place icons, connect them in a story, and bang the book is there” in a way that produces the required text output. That pushed the author toward a text-first pipeline.

The current stack is three tools. Gemini 2.0 (an experimental, free model) is used for large-context tasks; it can accept up to 2 million tokens in theory, though the author hit an internal server error above 1.4 million tokens. NotebookLM is used after the outline is drafted, drawing on previously filed content to identify missing elements and provide feedback grounded in reference materials. ChatGPT (paid, $20/month) is added to overcome free-tier rate limits and to take advantage of “projects,” which allow grouped project chats, document attachments for context, and custom instructions (target audience, intended style, and book purpose) to keep outputs consistent.

The refinement loop runs about five cycles. First, Gemini generates a rough outline from the author’s accumulated materials (transcripts, blog posts, and notes about visual PKM, plus selected books and articles). Then NotebookLM recommends what’s missing and where the outline needs strengthening. The updated outline goes back into Gemini for a second version. After that, the author tightens the structure by selecting reference books and using NotebookLM to generate a separate 20–30 minute “podcast” summary for each book. While listening, the author uses ChatGPT to produce dialogue-style transcripts that connect each source back to the outline and the author’s own experience. Those transcripts and the outline feed back into ChatGPT for targeted feedback, and the outline is edited in Obsidian. The result is an outline that’s not just derived from sources, but shaped by personal anecdotes and reflections.

Once the outline is robust, the chapter-writing phase begins. The author condenses the long outline so it fits into context limits and then generates chapters iteratively: NotebookLM supplies chapter-specific references, and ChatGPT is prompted to ask clarification questions first, then draft the section. The author reviews and edits the output in ChatGPT and saves each chapter into Obsidian. With the outline already refined through multiple cycles, the author expects that—if time allows—an entire rough draft could be ready by the end of Saturday, followed by fact-checking, reference verification, and adding illustrations.

Beyond the mechanics, the author is also soliciting feedback on cover designs and titles. The working title direction centers on “visual thinking for the digital mind,” with a related concept of an “ambidextrous digital mind,” and the author plans to choose a cover based on audience votes while keeping the March 15 launch goal in view.