Learning to see

A follow-up to The Human Touch.

I need to tell you something. I wrote an entire post about AI craft and human judgment, and then realized I had no idea how to practice what I was preaching. Not for the first time.

You know the person who gives great advice about work-life balance while answering email at eleven at night, or who writes eloquently about mindfulness while scrolling Twitter during meditation? That is me with attention. I can wax lyrical about the importance of human wisdom and contextual judgment, and my actual day is messier. I am the person who opens a browser tab to check one thing and surfaces three hours later reading about fifteenth-century manuscript illumination. I start deep work sessions with noble intentions and then "quickly check" Instagram for what turns into an hour of context-switching. I write about the irreplaceable value of human attention while my own scatters like leaves in the wind.

So after posting The Human Touch I was left with an uncomfortable question. How do you actually develop the kind of attention that lets you work with AI rather than be led by it? How do you cultivate the judgment I had been writing about? That question sent me down a rabbit hole I am still tumbling through, and it feels too important not to share, even though I am absolutely not an expert. What I have found I am calling attention architecture, and it is becoming the foundation for how I think about staying meaningfully human in an AI world.

The meerkat's profound wisdom.

Here the research got personal in a way I had not expected. Studies of meerkats (I know, I know) show that their sentinel behavior is not only about individual survival. They are more likely to stand guard when vulnerable members of the group are present; the vigilance serves the collective. And they have mastered something I keep coming back to: the ability to adjust their cognitive aperture to what the situation demands. A meerkat can scan the horizon for threats while staying alert to the family group below, shifting between narrow focus on a particular danger and broad awareness of the whole landscape.

This connects to something I touched on in the AI craft post without fully grasping it. The most valuable human contribution is not competing with the machine on raw processing but bringing distinctly human capacities, contextual judgment and collective wisdom, to work the machine augments. What I missed is that those capacities do not just happen. They need an attention architecture, the deliberate adjustment of aperture to task. The meerkats get it instinctively. I am learning it the hard way.

My messy experiments.

I want to be honest about what I am trying, because the internet has enough confident experts telling you what works. I share these not because they are proven but because the process has been valuable, even when I fail at it, which is often.

Morning horizon scanning (success rate: maybe 30 percent). I try to start each day with fifteen or twenty minutes of wide-aperture time: industry news, broader trends, long-term goals, looking for patterns and weak signals rather than reacting to whatever is loudest. Most mornings I still dive straight into email because it feels urgent and important. When I do manage it, the difference in the day is noticeable. I make better decisions about where to spend my time and catch opportunities I would have missed.

Managing input variety (success rate: wildly inconsistent). This may be my biggest challenge. I consume a strange mix of podcasts, one day a series on Byzantine history, the next someone explaining that AI coding is making programming languages obsolete and plain English is the new Python. The difficulty is knowing when the variety is useful cross-pollination and when it is just intellectual scatter. Sometimes the contrast between ancient history and new technology sparks a real insight about human patterns; sometimes it leaves me floating on the surface of everything and diving into nothing. I am experimenting with what I call thematic seasons, a few weeks on related topics before deliberately shifting domain, but I am still working out when I need more variety and when I need to resist the novelty and go deeper.

Transition rituals (success rate: about 40 percent). Between kinds of work I try a brief pause to adjust the settings. Before a strategy session, a few breaths and a widening of attention to take in the context. Before technical work, a deliberate narrowing to the problem in hand. It sounds simple and is surprisingly hard to remember; I catch myself switching contexts abruptly and realize I forgot to change lenses again. When I do remember, the quality of the work improves markedly.

The two-minute check-in (success rate: rising slowly toward 50 percent). Before any significant task, pause and ask what aperture it needs. Am I scanning for opportunities, focusing on execution, or trying to hold several streams at once? The failures are instructive. When I skip this I usually end up using the wrong kind of attention, scanning broadly when I need to focus, or diving deep when I need peripheral awareness.

Permission not to know yet (success rate: getting better). One practice that has been surprisingly liberating is allowing myself to say that I do not know enough to go deep yet and need more time at wide aperture. This goes against every productivity instinct I have, but the feeling that I must produce something rarely delivers the best results. Only a good deadline does that. Sometimes the most productive thing is to admit I am still in the scanning and absorbing phase rather than force myself into execution. It feels unproductive, and it often leads to much better work when I finally focus.

What it does to working with AI.

This attention work is reshaping how I think about AI collaboration. Rather than treating the machine as a more sophisticated search engine or writing assistant, I am starting to see that different AI work needs different human attention. Wide-aperture work, using AI to scan trends, generate perspectives or explore possibility spaces, needs broad awareness; too focused, and I miss the interesting tangents. Narrow-aperture work, refining ideas, working through structured problems, the occasional bit of code, needs sharp attention to the details and context only I fully understand; the machine does the lifting, I hold the standard. And there is a hyperfocal setting I am still learning to find, where AI is a thinking partner and I am focused on the immediate problem and aware of the wider implications at the same time. It asks for several levels of attention at once, and it is powerful when it works.

The difference is stark. When I am unconscious of my attention state, AI interactions feel mechanical and unsatisfying. When I match attention to task, the machine becomes a real thinking partner rather than a fancy tool.

The collective intelligence wake-up call.

This is where it connects back to the craft post, and where I saw how much I had missed. Research on collective intelligence shows that the smartest groups are not simply collections of smart individuals; they are marked by specific interaction patterns, how well members build on one another's contributions, keep diverse perspectives alive and balance individual insight against collective wisdom. But that only works if the individuals can manage their own attention first. If I am scattered, reactive and unaware of my own cognitive state, I cannot contribute to a collective intelligence, whether it is made of people, machines or both. I had been thinking of AI craft as an individual skill. It is also the capacity to take part in hybrid human-and-machine intelligence, which asks for a sophistication of attention I am only beginning to understand.

A few thinkers are reframing this for me. Jenny Odell, in How to Do Nothing, argues for resisting the attention economy not through productivity optimization but by reclaiming ownership of attention itself; life, as she puts it, is more than an instrument and so not something to be optimized. That helped me see attention architecture as something other than another productivity hack. Matthew Crawford, in The World Beyond Your Head, makes attention a moral and political question, and his insistence on engagement with the world of flesh, blood, wire and mud is a useful counterweight to AI's abstracted intelligence. Cal Newport's deep work and slow productivity, fewer things done better, fits with all of it; as AI takes over the shallow work, the people who can do deep, meaningful work become more valuable. The uncomfortable part is that these are not just pleasant ideas. They describe capabilities I need to develop to stay meaningfully human in an AI world, and I am not there yet.

Where I am going next, with realistic expectations.

I expect my understanding to change a good deal. A few directions feel promising. Experimenting across the spectrum, since some people love structure and frameworks and others prefer intuitive, meditative approaches, and what is emerging for me may not work for you at all. Understanding the stakes, because AI's abilities in pattern recognition and even creative tasks are advancing faster than many expected, which makes this more urgent rather than less; the question is not whether machines will get better at imitating human capabilities but whether we will develop the ones that remain irreplaceable. And building some company, because the response to the craft post was quiet but thoughtful, a few people wrote privately to say it resonated, and I suspect there are others wrestling with the same questions. Honestly, I need the help.

If any of this resonates, experiment alongside me. This is not about finding the right system; there probably is not one. It is about noticing your own attention patterns and being more deliberate about where you spend this precious resource. Start small. Before your next important task, ask what kind of attention it needs: wide, for scanning and connecting; narrow, for focused execution; or the hyperfocal sweet spot where you can hold several streams without losing quality.

I am convinced that learning to architect our attention is not only about personal effectiveness. It is about staying irreplaceably human while using extraordinary tools, and about preserving our agency in a world where it is increasingly easy to let the algorithms decide. As a Gen X man who grew up believing in the promise of technology and now watches it reshape everything in ways that are not always comfortable, attention architecture feels like a way to stay intentionally human while still engaging. It is not a rejection of AI, or nostalgia for simpler times. It is developing the capacity to choose how we engage rather than being carried along by whatever captures us most effectively.

I do not have this figured out. I am probably failing at it more often than succeeding. But I am learning, and that feels like progress. If you explore your own attention architecture, tell me what you find, especially the failures, because I suspect that is where the real learning is. The goal is not perfection. It is conscious evolution, stumbling forward together toward something more intentionally human.

This post represents my thinking, and my struggles, as of June 2025. I expect both to evolve, and I will share what I learn. If you are experimenting with this yourself, or have research or frameworks to point me to, please get in touch; this feels like work that benefits from collective intelligence, which I am still learning to contribute to.