A writer who has published 200 analyses of AI and mental health, and discussed the subject on CBS's 60 Minutes, is arguing that mindfulness and mindlessness are not states a person switches between but two processes sharing one mind in shifting proportions — and that generative AI is the instrument for adjusting the ratio. The claim sounds modest and is not. If the mix is continuous rather than binary, it can be observed, nudged and managed, and the essay proposes that ChatGPT, GPT-5, Claude, Gemini, Copilot and Grok are the systems to do it. A follow-up piece is promised on the prompts required.
The definition in use is not the one from meditation apps. It comes from Ellen Langer: mindfulness means refusing automatic behaviour and actively looking for new distinctions and features in a particular situation. Mindlessness is the opposite posture — stuck in prior categories, blind to what is novel in the surroundings, running a rigid and familiar script. It is a relatively passive state, and the author's point is that it is easy to fall into.
Earlier installments in the series set out five ways generative AI can build the mindful side. It pushes a person to evaluate their own decisions deliberately, revisit personal biases and look again at settled views. It supports metacognition — reasoning about how one is reasoning. Through conversation and different AI personas, it surfaces positions a person had not considered. It sustains intellectual curiosity and widens openness to new knowledge. And it interrupts habitual trains of thought, returning attention to the surrounding world. The result can be a virtuous cycle: use the system to widen your thinking, become more mindful, and use it again. In that role the author casts AI as a psycho-educational tool, and notes that regular interaction helps hold attention and resist sliding back into old negative thought habits. He also names the obvious failure mode in a single sentence: a person can end up mindful only because the system keeps prompting them, which is dependence rather than development.
The mindless side has its own mechanisms, and the essay lists them without flinching. Analysis gets offloaded — why work out what is happening when the system will. Deliberation is skipped, with the pros and cons handed over. Intellectual passivity sets in, because the answers are usually correct and the urge to check them dies. Metacognition weakens as fewer decisions are made unaided, and with it the ability to reason broadly and see the whole picture. At the level of a population, the worry is gradual: a little more thinking transferred each week, and over months or years the capacity to think independently measurably erodes. The counterweight the author offers is that AI can also report that a person is stuck in a familiar mental groove, and that the notification alone can be enough to start the shift.
The structural move comes next. Treating the two states as mutually exclusive turns consciousness into a zero-sum game, in which one must win and the other yield. The alternative is a vessel holding two liquids that mix, with the proportions changing. The illustration is a man at his dinner table, eating the same meal he eats every evening, prepared the same way from the same ingredients, while reading a newspaper and thinking hard about how the news will affect his life and what he may have to change. The essay asks which state he is in. He is mindless: he does not notice the extra pinch of salt or that the food is a little warmer than yesterday. He is also mindful: he is reading, and searching for new ways to alter his behaviour. The third reading is that both are running at once — say 5% of consciousness on the automatic eating and 95% on the reading and the thinking, a ratio that shifts from moment to moment.
That ratio is contested continuously. Put the paper down, concentrate on chewing and admit no new thoughts, and mindlessness takes the space. Put the fork down to chase an idea the paper started, and mindfulness takes it. The author's caveat is that the split rarely reaches 100 to 0 in either direction; 1 to 99 is more realistic, with a residue of the losing state always present. He also observes an asymmetry: a mostly mindful person can notice the automatic behaviour and deliberately reverse it — catching himself eating without tasting, then savouring each bite — whereas mindlessness rarely notices that mindfulness needs reinforcing. It simply expands whenever attention slackens.
From there, the regulation argument. AI is cast as a lighthouse that illuminates the current state of consciousness and steers it toward the mindful end, and five applications follow: detecting mindlessness, warning about it, helping a person track their own state, stimulating mindfulness so it takes the upper hand, and identifying a person's general leaning toward one mode or the other. Each will get its own treatment in the next piece, built around prompts written for ChatGPT, GPT-5, Claude, Grok, Gemini and other systems. The essay closes on Ralph Waldo Emerson: a mind stretched by a new idea does not return to its former dimensions.
The coexistence model is appealing because it matches how attention actually feels, and weak for the same reason — nothing in it can turn out to be false. The percentages arrive without an instrument. There is no account of what 95 to 5 would look like from the outside, or how it would be distinguished from 80 to 20, and the vessel metaphor is carrying weight that an operational definition would have to carry in any study designed to test the idea.
The harder problem is that the same system is cast as both the disease and the diagnosis. Read the two lists side by side: offloaded analysis, skipped deliberation, passivity toward answers that are usually right, weakened metacognition. That is a description of the exact interaction pattern the five regulation applications require. To have a model detect your automaticity you must route your attention through it; to have it warn you, you must accept its reading of your mental state rather than forming your own. Dependence is acknowledged in one sentence and then made load-bearing.
And the mechanism goes unexamined. A chatbot sees the text a person types, nothing else — not the dinner, not the fork, not the fact that the meal tastes the same as yesterday. The essay never explains how detection would work, and since the promised follow-up consists of prompts, detection appears to reduce to self-report: you tell the model you have been running on autopilot, and it tells you that you have been running on autopilot. That is not a sensor. It is a mirror with an encouraging tone, and "detect mindlessness" is the item the other four depend on.
What the coexistence model does deliver is a vocabulary for the argument about AI and cognition that is neither "it sharpens us" nor "it dulls us." That is more honest than either slogan. It is also the trouble: if the mix is always partial and always moving, then both camps are describing the same person in the same hour, and neither position has any state of the world left that could show it to be wrong.