Focus / Attention Mental Model (incl. AI vs Humans)
Executive Summary
“For a long time, we took our attention for granted, as if it was a cactus that would grow even in the most desiccated climate. Now we know it’s more like an orchid, a plant that requires great care, or it will wither. […]
I started to wonder what happens to a world where this form of deep focus shrinks so far and so fast. What happens when that deepest layer of thinking becomes available to fewer and fewer people, until it is a small minority interest, like opera?”
If you only have three minutes, the concept in brief: Attention is the “biggest flame” in our world today; collectively, our attention is as much in need of rescue as a company that’s a week away from a lender taking the keys. After the COVID era, which led to us all being online all the time rather than “in the real world,” I – like many others – found my attention span alarming. There were hours of my day I couldn’t account for, where I was neither at play nor work.
Attention underpins the success of both AI / LLMs and humans – but works in very different ways for each. Attention is the selective allocation of limited cognitive resources; focus is sustained attention toward a chosen goal. As one researcher described it, the prefrontal cortex acts as a “bouncer” at the door of your mind’s nightclub, deciding who gets to party. Unfortunately, that bouncer is now fighting off a veritable armada of outside forces – an environment of constant interruption and algorithmic harvesting that Johann Hari calls “stolen focus.” This problem likely to only get worse as AI reduces the marginal cost of content creation to ~zero. Managing attention (whether as an individual or an organization) is a critical prerequisite for success, particularly given that a substantial portion of current knowledge work can either be entirely automated or massively accelerated with AI.
Key takeaways/applications:
Stolen focus is structural. The modern attention crisis is not a personal failing but a systemic problem driven by the incentives (we’ll go deep into Johann Hari’s Stolen Focus). Everyone and everything (including this webpage) is constantly competing for your attention. Technology has been engineered to hijack our dopamine circuits; meanwhile, “always-on” work culture and RTO mandates create a work environment full of distractions. Willpower is not the solution; design beats discipline. To achieve success, we need to increase activation energy for attention sinks while managing attention like cash flow.
Context switching kills quality (and quantity). The “myth of multitasking” is a delusion; humans are single-taskers – more like CPUs than GPUs, as we’ll explore. As MIT’s Earl Miller explains, what we experience is rapid “task-switching,” which comes with a severe switching cost effect. If you switch tasks every few minutes, quality and speed both crater.
Design for your chronotype. A lot of productivity advice recommends specific times of day – often in the morning – to “get things done.” This is nonsense. Sleep deprivation is a massive attention killer, so the universal “5am club” is a myth; align deep work to your biology, not to a schedule that degrades health and performance for most people. (Internal Time by Till Roenneberg.)
Use AI to automate and scale tedious tasks, rather than outsourcing judgment. AI acts as an army of “infinite interns” that can automate tedious tasks (for example, the structuring / organization of this webpage – most of the writing is mine; most of the structure is AI.) Our brains are like CPUs (bad at parallel processing) while AI models are like GPUs (The Nvidia Way). This frees up your limited human (CPU) attention for the hard parts: synthesis, calibration, and decision-making.
Calvinistic productivity mandates are an incomplete schema. Regardless of the sort of work we do, many of us have a tendency to equate productivity with time; more hours on task = more work. But our brain at “rest” – or during sleep – is not doing nothing. Rather, “mind-wandering” or the default node network represent our brain trying to make sense of what we know and make connections – the sort of “lateral thinking” that AI simply cannot handle, but that leads to innovation and creativity. Of course, it’s garbage in, garbage out – if we feed our brain short-form junk like TikTok or memes, there’s no raw material to make interesting connections between.
Prefer to listen to this as a podcast? Find Poor Ash’s Almanack on Spotify, Amazon Music, YouTube, or Apple Podcasts. Or just watch below:
“Attention Is All You Need” – A Deeper Look At Attention and Focus
Attention underpins three critical trends that are shaping the modern world. We’ll start by briefly covering each, then go into more depth on the solutions.
1. Humans: A Century-Long Decline In Attention Spans
Like human memory, human attention is inherently constrained by biology. However, it’s a perfect example of the nature vs. nurture mental model. Unlike memory – where our inherent capacity is relatively limited (i.e. the impact of nurture vs nature is small) and the utility, or payoff of working on it is very low (since technology is simply better at remembering things than we are – even champions in professional memory competitions use grocery lists or forget where they’ve put their keys).
Conversely, there is a wealth of data demonstrating that human attention can, in fact, be cultivated, and that the effect of nurture rather than nature is enormous. It’s not just humans; either. Clive Wynne argues in Dog Is Lovethat a dog’s “genius” at reading human pointing gestures is not an innate, evolved trait. Instead, it is learned. This was proven by studies of shelter dogs, who initially failed the pointing test but “learned to follow the gesture in an average of just ten minutes.” Even wolves learned to follow it. This learning happens best during a “critical period” for socialization. Anyone who’s had a dog knows that they quickly learn to pay attention to things that are important to them (it took our dogs merely days to learn the sound of a bag crinkling, and to start paying attention).
Unfortunately, we have been going in completely the wrong direction for a century; an alien landing on Planet Earth might conclude that we had a society-wide initiative to completely destroy our attention.
Collective attention is measurably shrinking over both the long term and the immediate short term; the consequences are piling up across comprehension, memory, creativity, empathy, and even safety. As attention spans compress, we sacrifice depth, which “takes time” and “reflection.”
One analysis found that top Twitter topics stayed in view for 17.5 hours in 2013 but only 11.9 hours by 2016; broader analyses including Google Books data going back to the 1880s showed that “with each decade that passed, for more than 130 years, topics have come and gone faster and faster.”gThe same trend shows up in other sources of data, ranging from movie ticket sales to Google searches.
If you feel like the pace of change is accelerating, you’re not wrong. The pace of information consumption has accelerated; we are ‘drinking from a fire hose.’” Research shows that daily information consumption rose from “the equivalent of 40 newspapers” in 1986 to “174 newspapers” by 2007. Obviously, things have not gotten better for our poor brains since 2007; an obvious corollary is that the more information we are consuming, the less attention allocated to each. In Stolen Focus, Johann Hari quotes the work of research Sune Lehmann, which found “one mechanism that will [degrade attention spans] every time: You just have to flood the system with information.”
Our biological attention can be though of as a feedback loop, similar to training muscles via weightlifting. The more we use our attention, the better it gets; the less we use it, the worse it gets. Stolen Focus notes “the average American now spends seventeen minutes a day reading books and 5.4 hours on their phone.” Hari also notes that the average American touches their phone 2,617 times every 24 hours (a clear example of the activation energy mental model – they’re always present). Worse yet, much of this data hails from before much of social media moved to even shorter-form junk (such as TikTok or “reels” rather than actual videos).
Hari cites Professor Anne Mangen’s research, which shows that the medium of reading trains our brain. “Reading from books trains us to read in a particular way – in a linear fashion.” In contrast, reading from screens “trains us to read in a different way – in a manic slip and jump.” We “‘scan and skim’” on screens, and this becomes a new habit.
Rather than getting enough sleepand cultivating our attention span, we’re often trying to medicate away our exhaustion – and energy drinks aren’t the worst of it. ADHD diagnoses in the United States have soared; a high single digit percentage of girls and low-teens percentage of boys today are diagnosed with ADHD, with rates as high as 30% in some areas of the American South. The first-line treatment is often stimulants like Ritalin that are chemical analogs to methamphetamine – yes, the same crystal produced by Walter White in Breaking Bad – which don’t improve test scores much more than behavioral guidance, but creates substantial side effects.
This couldn’t come at a worse time: while estimates vary, many believe AI can potentially automate 60 – 70% of current knowledge work over the next few decades, depending on the field and task (some things may be as low as 20-30%, while others can be up to 90-100%). This is real, and here, not some far-off theoretical benefit; my Q2 2025 investor letter, “10x,” explains how we’ve increased our productivity by a factor of ~10x by leveraging AI. (Indeed, though most of this page is human written, you wouldn’t be reading it without AI). Fortunately, human attention and AI attention work differently, and by understanding the nuances of each, we can leverage our uniquely human strengths and have AI serve as a force-multiplier (rather than competition).
2. Attention in AI: Why Transformers Work (And Don’t)
The 2017 paper Attention Is All You Need sparked the AI revolution by changing how machines pay attention – introducing a new attention mechanism that underlies large language models like ChatGPT and Gemini. Instead of reading word‑by‑word (like recurrent neural networks previously used for deep learning), a transformer uses self‑attention to look across a whole chunk of text at once and assign learnable weights to what matters. In practice, that means it can connect “it” back to “animal” across a sentence and catch long‑range patterns humans often miss when reading linearly; this is hugely useful for translation, summarization, and retrieval at scale. Transformers process tokens in parallel, not sequentially, and that difference shows up everywhere from their strengths (speed, coverage) to their blind spots.
Hardware is the other half of that story; biological (human) hardware and computer hardware have analogies, but also differences. Transformers run on GPUs: thousands of small cores doing matrix math in parallel. Our brains behave more like CPUs: great at rich, serial, complex reasoning, terrible at running a thousand light, independent calculations at once. (See Tae Kim’s The Nvidia Way for more.)
But none of that means AIs “think.” As I argued in my Q4 2024 letter, LLMs are pattern engines: they predict the next token based on statistical associations, not an internal grasp of causality or truth. That’s why they’re shockingly good at benchmarks and boilerplate, yet still struggle with simple negation (“not X”), confound sarcasm, and can hallucinate confident nonsense when data is thin. They can simulate a chain of thought; they do not have one. Which is fine, as long as you use them accordingly.
Humans can excel at things AIs can’t – such as deciding where attention is valuable. In particular, humans excel at lateral thinking. When using AI – say, to code a custom Python application or solve a WordPress problem – I’ve found that they often display commitment bias of a sort, getting stuck on a certain path. A real example on Poor Ash’s Almanack: I wanted to figure out a way to embed a podcast player into the webpage, but most of them either required a paid subscription to the widget, or required you to log in (such as Spotify). Despite multiple queries, neither ChatGPT nor Perplexity came up with the obvious answer that I did: embed a YouTube video (since the podcast in question is also hosted on YouTube). The AI got so focused on the idea of a “podcast widget” that it failed to consider other approaches that accomplished the same goal (allowing any visitor to the website to listen to the podcast, without requiring sign-in on thir end or a paid subscription on mine!)
This highlights the shortcomings of thinking of our focus or attention only through the lens of a “spotlight” – an example of the product vs packaging mental model, discussed in various ways in Johann Hari’s Stolen Focus, Alex Soojung-Kim Pang’s Rest, and Steven Johnson’s “How We Got To Now.” Breakthrough ideas rarely arrive from grinding harder; they come from the brain’s mind-wandering “default mode network (DMN)” weaving distant associations during apparently “idle” time, such as walks, showers, and the drive home (anytime when the brain is not otherwise cognitively occupied). Critically, as we’ll look at in a bit more depth in the next section, the DMN will not activate if we are infinitely scrolling through TikTok – this thus serves as a double-whammy negative, not only cluttering our memorywith useless junk, but also reducing the amount of waking hours we spend making connections (and, possibly, sleeping hours, if we’re on our phone in bed.)
Hari’s Stolen Focus pulls together research showing that mind‑wandering correlates with memory, imagination, empathy, and meaning; it’s often when creative leaps happen. Pang’s Rest makes the same point: deliberate rest is a productivity tool, not a luxury. Transformers don’t have a DMN. That’s why AIs are bad at lateral thinking in the human sense – and why our edge is often making unexpected connections, or judgment more generally.
Steven Johnson’s work in How We Got To Now powerfully reinforces this. The “distant associations” from the DMN are exactly what he calls “slow hunches,” which he identifies as the true engine of innovation. Johnson coined the term to describe the innovative process where an idea takes shape over years through gradual development rather than a sudden flash of insight (and certainly not through the Edison model of invention we all think of, i.e. a scientist tirelessly toiling away in a lab trying hundreds of combinations with direct, focused effort).
The book is a catalog of these creative leaps. For instance, Willis Carrier’s “accidental discovery” of air conditioning: he was hired to solve a humidity problem for a printing plant and realized the machine that removed humidity also chilled the air, a lateral insight that spurred a massive demographic migration to the “Sun Belt” that continues today. Similarly, Clarence Birdseye’s “slow hunch” connected his experience with Inuit flash-freezing fish in -40 degree air to the mushy, inferior quality of the slowly-frozen food of the day, thereby creating the modern frozen food industry.
3. Are You Still Reading? The Attention Economy
For any business, attention is the ultimate precursor to revenue: if you don’t notice the business exists, you’re hardly likely to spend there. Businesses have always tried to capture our attention through billboards or witty advertising, which means that advertisers have always had a keen interest in driving engagement. Indeed, decades before personal computers (let alone smartphones) existed, the “father of American Advertising” David Ogilvy was hunting out “factors” (i.e. base rates) that make audiences respond better or worse to print advertisements. (See The Unpublished David Ogilvy or Ogilvy on Advertising.)
Of course, when the incentiveto make money is married with technology so advanced it’s indistinguishable from magic, real problems can begin. In Stolen Focus, Johann Hari explains in great depth how when there’s no product… you are the product. For social media platforms, the longer your eyes stay on a feed, the more ads they can show and the more data they can mine, so it’s designed to keep you there with tactics like infinite scroll (which reduces the activation energy of consuming more content to zero). COVID made all of this worse; for months (or years), we were forced to live exclusively through screens; not only were restaurants closed, but in the early days, sometimes even parks and beaches. This created a habit that created a feedback loop.
Worse yet, ranking algorithms promote content not based on intrinsic quality or utility, but rather what gets more clicks – often things that push people towards extreme ideology. This creates an arms race or feedback loop – as attention spans shrink and people’s preferences shift more towards bite-size content, more such content will be created, which reinforces the trend towards shorter attention spans (since nobody is training their attention muscle).
As individuals, we should push back, and direct our focus and attention to what is really important. As Hari puts it:
What is th[e] message [of Twitter?] First, you shouldn’t focus on any one thing for long. The world can and should be understood in short, simple statements of 280 characters. Second: the world should be interpreted and confidently understood very quickly. Third: what matters most is whether people immediately agree with and applaud your short, simple, speedy statements. A successful statement is one that lots of people immediately applaud; an unsuccessful statement is one that people immediately ignore or condemn. When you tweet, before you say anything else, you are saying that at some level you agree with these three premises.s You are putting on goggles and seeing the world through them.”
That last bit is a perfect example of the schema mental model. Of course, the problem isn’t just social media; even if we didn’t use social media, we would still be distracted by the mere presence of our phones. Smartphone apps use similar tactics such as streaks or push notifications. All of this will get worse as algorithms get better, particularly as AI drives the marginal cost of content toward zero, and can process infinite numbers of A/B tests to see what drives engagement.
More speed means less comprehension. When feeds teach us to skim, we become less willing, and less able, to grapple with complex material. The DMN doesn’t get the white space it needs to knit ideas together, so lateral insights never come up. We’re stuck in what Hari calls “an unsatisfying whir… we’re not focusing, but we’re not mind‑wandering either.”
A key problem is rapid context switching. The idea of “multitasking” is a myth that originated from the computer concept, but our brains are not computers. We don’t multitask; we switch tasks. Hari quotes blunt commentary from MIT’s Earl Miller: your conscious mind has a bottleneck; it can hold one or two thoughts at once, full stop. Every switch carries a switching cost (error + backtracking + re‑orientation). In controlled settings, intermittent texts cut student performance by ~20%; routine digital interruptions drop measured IQ by ~10 points.s That’s twice the impairment of marijuana, and not dissimilar from the IQ delta from an engineering graduate to a member of the general population who might not even have a college degree. In other words, attention fragmentation literally makes you stupider.
Going back to the CPU/GPU analogy, think of our brains as CPUs: excellent at sequential, high‑context reasoning, judgment, and sense‑making, terrible at repetitive, parallel grunt work (such as data entry). AIs, or technology in general, are GPUs: massive parallelism that excels at extensive, shallow computation (token‑wise patterning), but are brittle at deep causal judgment. Both have strengths; both have ceilings. The mistake is trying to treat your brain like a “GPU” – i.e. trying to process many things in parallel. Of course, treating your AI like your brain – expecting it to supply meaning – is a mistake of equal proportions.
Speaking solely for myself, I reached a point where my fragmented attention was crisis number one: I was wasting precious hours of my life without meaning to, existing in that “unsatisfying whir” – neither productively working, nor spending time on hobbies I truly enjoy, or with my family (including our two sweet Labrador mixes). I realized that I needed to view attention and focus through the lens of addiction and habit – the same dopamine circuits are at play.
If you feel this way, or if – like Hari – you have an adolescent in your family who is buried in their phone – the consequences are simply too high to ignore. It’s not “okay” or something they’ll grow out of. Attention span should be the number-one priority of parenting in this generation (I’ll get into some of the data on kids in a subsequent section). Regardless, after realizing my attention span had been fried by the COVID era, I have – over the last few years – started focusing on attention as a core principle, the same way corporate turnaround expert Jeff Sands approaches resuscitating a business on death’s door:
“Biggest flames first. That’s it, you’re going to be firefighting until you can get the business stabilized. Forget strategy, forget the income statement, forget whatever you might have had planned for the next 100 days. You’re firefighting.”
Solving The Attention Crisis
Stolen Focus is a great book; the problem is that Johann Hari is a socialist. Like all socialists, he underweights agency (free will, or “belief behavior matters” – our own individual choices to make of our lives what we want of them). He makes the same mistake with the attention crisis that he made with drug addiction in Chasing The Scream: it is entirely within each of our individual control to address these challenges.
What I’m not arguing is that willpower is the answer; willpower is almost never the answer (willpower and agency are very different concepts). We’re working against systems that have used choice architectureand activation energy to make it easy to not pay attention to what we should really be paying attention to. We have to do the same, in reverse. Indeed, Stolen Focus quotes Roy Baumeister, a leading expert on willpower, who confesses his own struggles.
Baumeister states: “‘I’m feeling like my control over my attention is weaker than it used to be.’” He explains that he used to read for hours, but now “‘it seems like my mind jumps around a lot more,’” and he finds himself turning to phone games when he feels bad, “just sort of giving in to it.”
This highlights the challenge of maintaining willpower in the modern environment – even for someone who has studied it for more than thirty years!
Make Technology Work For You – Don’t Work For It
To be clear, this is not intended to be a “productivity guide” – there are plenty of those out there. This is intended to specifically focus on how we can protect our focus and attention, which has the natural corollary of being more productive – but is simultaneously broader and more specific in nature.
Technology Is Here To Stay
As I’ve become something of an AI evangelist – a strange position for me, as I’m generally a skeptic of the current “hot thing” (I still can’t understand the existential purpose of a Bitcoin, and no, please don’t kindly email me an explanation) – I’ve been surprised by the level of pushback from many otherwise-smart people I know. I believe I’ve identified at least two reasons for this.
The first is being uncomfortable with tradeoffs or incorrectly fearing n-order impacts from technology.
Don Norman rightly views the glorification of memorization in the educational system and elsewhere as immature and inappropriate, delivering a phenomenal analysis toward the end of The Design of Everyday Things. Every technology – including books (which we are now nostalgic for, in the age of tweets) – faced skepticism. He explains:
“In Ancient Greece […] Socrates complained about the impact of books, arguing that reliance on written material would diminish not only memory but the very need to think, to debate, to learn through discussion […] but over the years […] human intelligence has certainly not diminished.
Does the fact that I can no longer remember my own phone number indicate my growing feebleness?
No, on the contrary, it unleashes the mind from the petty tyranny of tending to the trivial and allows it to concentrate on the important and the critical.
[…] the power of the unaided mind is highly overrated.
[…] Human intelligence is highly flexible and adaptive, superb at inventing procedures and objects that overcome its own limits. The real powers come from designing external aids that enhance cognitive abilities.”
Indeed, Stolen Focus has some advice that I don’t agree with (even beyond the socialist nonsense). For example, page 82 highlights the concept of “screen inferiority,” which is supported by “broad scientific” evidence from fifty-four studies. This research shows that “people understand and remember less of what they absorb on screens” compared to paper. This comprehension gap is significant; in children, “it’s the equivalent of two-thirds of a year’s growth in comprehension.”
I don’t doubt that these findings are real; conversely, given the speed, breadth, depth, and cross-referencing that technology can enable when used well, I have zero intention of having my own kids work very much with pen and paper. (I was homeschooled, and we plan to homeschool our kids, so I think a lot about education; I’ll get to this a bit more later.)
As a knowledge worker, the idea of using paper for any sort of serious work is laughable – not being able to use Ctrl+F or Ctrl+C, and of course now not being able to query against massive databases with NotebookLM or fact-check statistics with Perplexity, seems patently absurd. I love reading long-form books in depth to learn about a topic and benefit from an author’s expertise, but I would be doing my clients a vast disservice if I printed out and read an annual report on paper.
An easy antidote, of course, is to use the right device for the right topic. For work, computers are generally better than tablets are better than phones; simply not using your phone whenever possible is helpful. Phones are extremely convenient (they lower the activation energyof being distracted).
I think the second could simply be a bit of head-in-the-sand – if you can’t see AI, then maybe it won’t take your job. As Carole Hooven discusses in T: The Story of Testosterone, our feelings or concerns about a topic are “irrelevant to the truth of scientific hypotheses.” She warns:
“if you find a hypothesis distasteful, a red flag should immediately go up: there is a clear and present danger that you will discount the evidence that supports the hypothesis.”
Technology, including AI, is powerful and here to stay; rather than clinging to outdated and antiquated notions (such as my mom bemoaning my poor handwriting in 5th grade – neat handwriting is no longer important and has never once been relevant professionally), we need to accept that we are in an era where AI will have increasing impact in our lives, both directly and indirectly.
For example, I’ve recently seen some Twitter posts (yes, I still use Twitter a bit – I’m only human) that claim AI writing is low-quality and go so far as to classify all AI writing as “slop” and claim that nobody should use AI for business writing. Speaking as someone who has used AI extensively for both internal research documentation and writing portfolio commentary for clients, this is a point of view that can only be described as “objectively dumb.” It’s not that hard to improve the quality of AI writing (my wife has a nice guide on our sister site, Ravensight AI, on how to use AI for business writing), and by combining rapid AI summarization and structuring with human editing and insight/judgment, I can document my own research faster, and provide our clients with more transparency – all while freeing up more time to reinvest in better results.
So, we might as well leverage the power of AI rather than letting it leverage us, by focusing on what we’re good at and not getting sucked into the noise. We’ve all gotten used to using calendar apps to keep our appointments or grocery lists to help us make dinner; unless we’re talking about a mere handful of items, few people seriously try to memorize their own schedule or what they need from Costco.
The good news is that humans are adaptable. In addition to history, this is a core feature of our biology. We discuss this in more depth in the nature vs nurture mental model, which is closely tied to trait adaptivity; Marlene Zuk’s Paleofantasyand Carole Hooven’s T: The Story of Testosterone are good examples. Zuk notes that it’s a myth that there was ever a single, static “ancestral” environment we were designed for; the life and diet of an Inuit was vastly different from our ancestors on the hot savannahs of Africa.
Evolution never builds perfect organisms; it tinkers and creates “good enough” solutions that are riddled with compromises and tradeoffs; natural selection is described as being “more like a ‘broken zipper’ than a master engineer – but there’s a lot of room for variability. For example, Hooven discusses the very clear base rates on crime: in every country in the world, men are responsible for the overwhelming majority of violent crime. As she explores in a very nuanced way, men are simply more predisposed to (and capable of committing) violence, but that doesn’t mean that biology is destiny – due to different culture, crime rates in places like Singapore are very different than they are in the U.S.
The AI Angle: Outsourcing to AI to Prioritize Creativity, Judgment, and Deep Work
The solution to technology-induced technology problems is, in fact, technology. In Stolen Focus, Hari notes the following:
“We are, collectively, experiencing a more rapid exhaustion of attention resources… [when I detoxed], I realized… I was, for the first time in my life, living within the limits of my attention’s resources. I was absorbing as much information as I could actually process, think about, and contemplate – and no more. The fire hose of information was turned off. Instead, I was sipping water at the pace I chose.”
Of course, this was during a full digital detox, which is simply not practical on a daily basis (although perhaps helpful for the occasional vacation). Particularly intriguing in a more day-to-day context is Hari’s intriguing
He cites Aza Raskin, the regretful inventor of “infinite scroll,’ who explains how social media sites could be technically redesigned to be less harmful, using choice architecture. For instance, Facebook could “batch” notifications, so as Hari puts it, “instead of getting ‘this constant drip of behavioral cocaine,’… you would get one daily update, like a newspaper.” Raskin also suggests turning off infinite scroll (his invention) which “catch[es] your impulses before your brain has a change to really get involved and make a decision” so that users must make a “conscious decision to carry on” rather than being passively and endlessly pulled into more content.
What’s interesting is that the same technology (AI) that can be increasingly leveraged by big tech companies to distract us, can be used by us to protect our own attention. It is not only possible – but actually trivial – to build AI automations that can do exactly what Raskin and Hari recommend above, i.e. “batch” notifications so that you get an update in one chunk at predetermined intervals – rather than sucking you into your inbox or Slack tab, where you get caught in the deluge. It can also be used to build simple custom Python scripts that handle rote daily tasks (so that things are accomplished with machine precision, or blending logical, always-consistent code with the useful interpretation of LLMs).
AI automations can be built with urgent keyword alerts, such that you do get notified for truly urgent matters, but anything that can wait… will wait. (My wife has a helpful guide on our sister site, Ravensight AI, on how you can get started building these yourself – for free – and can also help any interested individuals or small businesses who want to outsource it. And yes, she eats her own cooking, using these exact automations to run her own business! Demo video below.)
I’m fortunate, by the nature of my work (as the solo manager of a small-cap investment firm), to not have to deal with many of these issues; my work, by its nature, is generally quiet, solo, and reflective. Nonetheless, I’ve built many other AI automations to handle small, rote, tedious daily tasks (such as renaming and categorizing downloaded files, or extracting data). Doing so has a double benefit – of course, it saves me time in the absolute sense, but more importantly, it saves my focus andattention – recall the earlier discussion about the quality, speed, and even IQ costs of context-switching.
Interruptions kill focus: Stolen Focus references a Carnegie Mellon study showing that students who received intermittent texts “performed, on average, 20 percent worse,” while simple digital interruptions produced an “average drop in the workers’ IQ by… ten points,” a hit “twice” that of smoking marijuana (and close to the difference in IQ between engineers and the population median).
Flow is the counter‑state. When your attention locks onto a real challenge at the edge of your ability, the work becomes self‑propelling; you get lost in it. Mihaly Csikszentmihalyi established the concept after studying artists, rock climbers, and others who exert tremendous focus over many hours. Flow requires a clear aim, a challenge that isn’t trivial or overwhelming, and – the scarce ingredient now – an environment that doesn’t keep yanking your mind out of gear (like an open office, or a series of digital distractions).
Hari cites flow; in Deep Work – published a decade ago but perhaps more relevant than ever today – Cal Newport comes at the same target from another angle. Newport posits that deep work is the repeatable way professionals reach flow on demand.
Newport’s argument is simple, and merciless. Deep work is valuable, rare, and meaningful; shallow work is cheap, abundant, and loud. If you don’t draw that line explicitly, the shallow wins by default because it’s easier to start and easier to measure (similar to the “urgent, but not important” quadrant of the matrix in Stephen Covey’s 7 Habits of Highly Effective People).
The answer is to flip the default. Newport takes the structural problem solvingview of the world, arguing (as I do) that willpower is overrated:
“Your will […] is not a manifestation of your character that you can deploy without limit; it’s instead like a muscle that tires.
This is why the [studied subjects] had such a hard time fighting desires […] the same will happen to you, regardless of your intentions, unless, that is, you’re smart about your habits.”
Newport recommends deciding ahead of time when you will do deep work (there’s no specific ideal time – that depends on your biology and your lifestyle), and protect those hours. He provides different options that may work better or worse depending on the nature of your work – monastic (nearly all depth), bimodal (whole days or seasons carved out), rhythmic (the reliable, daily block), and journalistic. The point is repeatability.
Deep Work also cites “attention residue” as a silent saboteur (similar to the statistics cited above from Stolen Focus). Every peek at the inbox throws sand into the gears; even when you return to the main task, part of your head is still chasing the other thread. That cost accumulates quickly. Newport’s fix is an example of structural problem solving: batch the shallow into small, explicit windows (easier now than ever with AI, such as the “digest” approach I mentioned above).
Newport also brings up an interesting concept – cultivating boredom (he has an entire chapter about it). His core point or observation is that if you flood micro‑gaps with novelty (checking your email, or social media, during every commercial of a TV show or scrolls) then you teach your brain that discomfort ends the second it begins, which makes it harder to resist the mind-wandering that naturally occurs during work.
Newport also recommends leveraging commitment bias on occasion by making a grand gesture such as a costly room or weekend lock‑in for a stubborn project, similar to Hari’s discussion in Stolen Focus of the “Ulysses” approach (tying yourself to the mast).
The real psychological keys here are twofold (and, certainly, I acknowledge that they are easier to implement if you are a principal rather than an employee.)
The first is the 80/20 principle and utility. The truth is that for many of us, most of our value does not correlate linearly to hours worked. Data cited in Stolen Focus (and corroborated by many other studies and surveys of the workplace) shows that many workers are only “engaged in their work for less than three hours a day.” When one New Zealand company proposed a two-month trial of a four-day week for five days’ pay, “[the] head of HR literally fell over.” But it worked. As one employee put it, “I was less likely to be distracted.”
The data backed it up. “Time spent on social media at work… fell by 35 percent,” and “stress levels were down by 15 percent.” Management “conceded that the company was achieving as much in four days as they had before in five.” Similar experiments elsewhere yielded success; although results will obviously vary and this is not practical for every line of work, it does demonstrate that it is possible to get as much done in less hours if you utilize focus.
The second is – going back to 7 Habits – proactively focusing on what’s important (not just what’s urgent). No matter what you do, it’s likely that at least 10-20% of it doesn’t need to be done – or could easily be outsourced to AI. We simply have to avoid FOMO and be comfortable with tradeoffs.
Stress, Sleep, and Diet
I’ll keep this brief because it’s relatively self-explanatory, but the research also demonstrates that other forces are damaging our attention. As Dr. Matthew Walker comprehensively overviews in Why We Sleep, exhaustion is the new normal. Sleep deprivation destroys the brain-cleaning process that removes metabolic waste during rest. If you wake up early (usual, but stupid and wrong productivity advice), you miss the last two hours of a normal night and you lose most of your REM cycle, which stabilizes emotions and cements learning. (See also Till Roenneberg’s Internal Time.) Sleep deprivation decouples the amygdala from the prefrontal cortex, producing overreaction and poor judgment. It also scrambles hormones, driving hunger for fast fuel while degrading nutrient processing. In other words, skip sleep and you don’t just lose focus; you biologically crave the very diet that erodes it.
Meanwhile, prolonged stress causes threat-monitoring narrows the lens to danger cues, which crowds out learning and long-horizon thinking. Johann Hari identifies this as “hypervigilance,” a state where the brain is constantly “scanning for unexpected dangers” (Stolen Focus). This isn’t a failure of attention; it’s an adaptive response to an environment perceived as unsafe. This is powerfully illustrated in Rob Henderson’s memoir, Troubled, where he describes his childhood in foster care as a “relentless ‘fight-or-flight’ state that precluded deep, contemplative thought.” When survival is the priority, “spotlight” focus on complex, abstract tasks is a luxury. Fortunately, for many of us, survival is not a real challenge, but our brain doesn’t know that.
Attention varies with environment. Hari discusses animal behavior across other books (such as “Chasing the Scream“). He notes that animals in captivity display behaviors such as pacing (or “cribbing” in the case of horses) that aren’t seen in nature. The problem isn’t the bear or the horse; it’s the “frustrated biological objectives” of its environment. Hari’s prior books (Chasing the Scream and Lost Connections) go into some interesting elements (such as our lost connection with nature and the meaning of our work) that are interesting to explore further; again, I think addiction is an interesting lens through which to explore attention, because I believe that a substantial majority of society is addicted to the daily dopamine rush of notifications.
Kids: At Greatest Risk
“You just did it for the dopamine. You didn’t mean to leave me so !@#$ed up. You did it for the dopamine. And could it be that I’m just not enough? You just did it for the dopamine. And it didn’t mean anything at all.
– “Dopamine” by Sum 41
Here’s interaction between attention and nature vs nurture. Since my wife and I are planning to start a family soon, we spend a lot of time thinking about mental models as they relate to parenting and our kids’ development.
I try not to speak in absolutes because the world is full of nuance – but I firmly believe that the research strongly supports that adolescents (let alone infants) should not consume any short-form or “junk” / hyperstimulating content at the critical points in their lives when neuroplasticity is highest and they are shaping their brain – and habits – for the rest of their lives. The consequences of damaging their attention are simply too high. My wife and I have real firsthand experience with seeing this play out; I’m choosing not to share the details for various reasons, but we’ve encountered someone whose brain was severely damaged by, essentially, TikTok and VR. It got better with a lot of focused effort (pun intended), and environmental modification, but ultimately the dopamine hits won out.
While many adults focus on “screen time,” not all screen time is created equal. TV shows, movies, or video games with a narrative or strategy can engage the senses and even build empathy; conversely, memes or over-engineered YouTube videos designed to hook kids’ attention (let alone TikTok) are literally degrading their intellectual capability in real time – the concept of “brain rot” has real scientific backing. This is of course hard for a parent to enforce in the broader culture and due to peer pressure (social connection with peers) – but the benefits in the long-term are well worth it; the effect of attention on measured IQs can approach a standard deviation – approximating the difference between the average engineering graduate and the population at large.
Then there’s the issue of ADHD diagnoses, which I touched on earlier. My mom has long believed that I would have been diagnosed with ADHD if I had gone to public school. I’m naturally a bit distractible, mostly because I tend to process information quickly and get bored easily with repetitive and/or non-challenging material (which is what elementary school is). I often played with Legos while learning vocabulary or spelling, to keep my hands occupied (something that wouldn’t be tolerated in school). My mom has said, many times, “little boys aren’t made to sit quietly in a classroom for hours on end.” She heard from friends’ parents about teachers recommending their kids be evaluated for or medicated for ADHD – which they did not have; they were simply smart, bored, and wanted to stretch their legs. I delve into this topic further – building in additional science from other sources – in the new nature vs nurture mental model.
I’ll leave off with a brief story about unintended consequences.
As I began to implement and use Askeladden’s new AI-driven research process, I felt like I had been handed a superpower. It didn’t just summarize; it cross-referenced, synthesized, and contextualized thousands of papers in minutes. A literature review that used to take three days was done by noon. I had successfully turned the “fire hose” of information into a civilized trickle. My logical conclusion was: I’ll finally have my evenings back.
The reality was the opposite, a perfect illustration of Jevon’s Paradox: the efficiency of the new tools increased my demand, consuming all the savings. That massive, newfound capacity (the time and mental bandwidth the AI freed up) didn’t translate into rest. It translated into an urge to maximize the new output potential. If I could now do three days of work in one, why stop after the first day’s worth? I began chaining together projects, feeding the AI new, complex queries after dinner, and then diving into the results. I started waking up in the middle of the night, my brain already churning through the next sequence of prompts, feeling low-level anxiety. Ultimately, my wife had to stage an intervention.
Cultivate and manage your attention as carefully as your capital; attention is essentially our intellectual capital.s When information is a “fire hose,” the solution isn’t to drink faster; it’s to be more selective about where you point the hose, and to use AI to take a lot of water out of the hose – so we can sip the amount we can handle.
In the era of AI, our human judgment and attention are our ultimate competitive advantage – we should use them wisely.
Conversely, there is a wealth of data demonstrating that human attention can, in fact, be cultivated, and that the effect of nurture rather than nature is enormous. It’s not just humans; either. Clive Wynne argues in Dog Is Love that a dog’s “genius” at reading human pointing gestures is not an innate, evolved trait. Instead, it is learned. This was proven by studies of shelter dogs, who initially failed the pointing test but “learned to follow the gesture in an average of just ten minutes.” Even wolves learned to follow it. This learning happens best during a “critical period” for socialization. Anyone who’s had a dog knows that they quickly learn to pay attention to things that are important to them (it took our dogs merely days to learn the sound of a bag crinkling, and to start paying attention).
