Independent Thinking in the Age of AI | PKM Summit 2026

AI’s biggest threat to thoughtful people isn’t job loss—it’s the erosion of personal agency: the risk that machines do the thinking while individuals drift into accepting AI-generated conclusions. The talk frames that danger through a simple question—how can people retain agency when prediction and generation tools increasingly supply answers before anyone has fully formed them? It ties the concern to a broader life problem: how to choose a next direction when success and stability leave room for doubt, echoing Dante’s restlessness and Steve Jobs’ warning not to live by other people’s thinking.

A central thread runs through everyday examples. Gmail’s sentence predictions raise a practical dilemma: when AI completes a thought, does the person still “drive,” or does the system steer the train of thought? The same concern intensifies when AI drafts entire emails—approval at the end may hide the fact that the underlying reasoning may not have been personally constructed. That leads to a second claim: text is a poor medium for how human minds work. Reading is linear and word-by-word, forcing the brain to stitch meaning together while constantly checking fit with prior knowledge. The talk contrasts that with visual recognition, where pattern detection can be near-instant—like spotting an “odd one out” in a diagram faster than a text-based AI response.

Evolutionary psychology supplies the rationale. Human cognition is portrayed as “navigation” rather than abstract contemplation: ancestors survived by tracking space, movement, and concrete cues, supported by mechanisms such as mirror neurons that simulate observed actions. From there, the talk argues that visual thinking creates a feedback loop: drawing an understanding lets people verify whether their mental model matches what’s on the page, making comprehension and questioning easier. Research on memory is used to bolster the point—concrete nouns tend to stick better than abstract concepts, and illustrations help convert abstraction into something graspable.

To make the case tangible, the talk uses two games. In the “game of 15,” adults struggle when numbers must be juggled mentally, but the same problem becomes trivial when the layout is visualized. A magic square demonstrates the same idea: representing the problem so structure is visible turns a hard search into an obvious pattern. A quote from Saul Emmerich frames the principle—solving means representing the problem so the solution becomes transparent. The talk extends this into a method: try multiple representations through iterative experimentation until a better tool reveals a path that wasn’t obvious at first.

The talk then targets “lazy brains,” aligning with System 1 thinking from Daniel Kahneman’s framework: fast, reactive pattern-matching that often jumps to conclusions. To counter it, it recommends injecting exploration between seeing and judging—collecting additional data points and using notes to overcome short-term memory limits. But notes shouldn’t be only linear text. Diagrams are presented as “visual prosthetics,” akin to switching from a Walkman’s linear playback to iTunes’ whole-library view.

Finally, the talk addresses a paradox: people naturally organize physical belongings when packing, yet default to text when organizing ideas. The proposed fix is to build a library of “visual primitives”—spatial layouts and diagram types that support thinking, not just decoration. It demonstrates this with circle maps and a double bubble map to compare perspectives in a real dilemma: quitting a day job to start a meaningful business. By mapping what each person values and fears, differences become explicit and empathy becomes structured.

The closing pitch is practical PKM design. Traditional note systems are described as one-dimensional (individual pages) or two-dimensional (folders and linear text), while modern linked-note tools still present a wall of text. The talk promotes “4D visual PKM,” where visuals lead and text becomes metadata—using tools and workflows centered on Obsidian and an “Excalidraw”-style visual-first approach. The goal is independent thinking: keep agency by choosing representations that make reasoning visible, verifiable, and personally owned.