Two Workflows for Reading, Note-Taking, and Visual Thinking that Are Transforming the Way I Use AI

AI use doesn’t have to mean handing over private notes. Zsolt’s two workflows aim to keep personal information “ring-fenced” while still using AI to generate summaries, connections, and visual structures that make reading and knowledge work stick.

The first workflow targets non-fiction reading, where audio-only often fails to produce recall. Instead of treating an audiobook as the primary input, he listens to the book on Audible while keeping ChatGPT open on his phone. After roughly 5–10 minutes (or whenever an idea hits), he pauses the chapter and records a short, interactive voice conversation with ChatGPT—positioning it less as a long chat partner and more as a reflective notebook. He uses the discussion to refine articulation (especially since he’s not a native speaker) and then asks ChatGPT for a concise summary at the end of each chapter. Crucially, he imports only that AI-generated summary into Obsidian—not highlights or raw book notes—then uses a “flipped notes” layout to pair the summary with an illustration created from the same conversation. He also keeps each chapter in a separate chat so ideas don’t bleed across chapters and so the workflow stays manageable.

Several practical rules make the method work reliably. Long voice reflections can trigger interruptions, but holding the screen prevents ChatGPT from responding until he releases it. Multitasking breaks the experience; he recommends staying physically engaged—walking, pacing, or even using idle moments like standing in line or waiting at traffic lights—because listening plus voice reflection doesn’t require constant attention on the phone. He also insists on timing: generate the summary immediately after finishing the chapter, import it right away, and move on. That freshness helps catch omissions or additions ChatGPT makes compared with what he personally considered important.

The second workflow builds a visual “external loop” from a whole body of knowledge. He starts by feeding books and notes about visual thinking into NotebookLM, then prompts it to generate a podcast-style overview (with a tight 500-character prompt limit). After downloading the podcast, he converts it to MP3 with Audacity, imports it into Obsidian, and uses NotebookLM again to transcribe the audio. Next, he uploads the transcript to ChatGPT to produce a Mermaid mind map; since Excalidraw struggles with mind maps, he requests a Mermaid flowchart instead. He then imports the flowchart into Excalidraw, rearranges branches into a more map-like layout, tweaks styling, and splits the canvas so the mind map sits beside the text while he creates illustrations. The result is a connected visual index: book covers act as links into his Obsidian vault, and the map becomes a jump-off point for future exploration.

Together, the workflows aim at selective sharing—external materials like books and curated reflections rather than a dump of private notes—while still producing summaries, structure, and new perspectives. Both workflows are presented as free to try: NotebookLM, ChatGPT, Obsidian, and Excalidraw are all used without paid subscriptions, with the broader claim that motion-friendly reading and AI-assisted visual synthesis can change how information turns into usable knowledge.