This is still early days for this. I’m in the middle of editing, I’m trying out a new way of writing essays with AI rather than doing it all solo, or all AI. There’s a LOT to do here, but we’re getting there.

---

We are busy in order to have leisure. — Aristotle, Nicomachean Ethics

A scenario:

A senior person is handed a difficult idea. They read the first page, skim the rest, and conclude that the thinking is muddled and the writing unclear, so its author should go away and make it clearer and simpler.

Sometimes that is the case, but often what is missing is not clarity in the document but attention in the reader, who has none to spare. One encounter cannot tell you which failed, the idea or the attention, but, over time, a distribution can. An idea that draws the same objection from many readers is probably the idea’s fault; one that only some readers cannot take might well be down to the reader. The manner of the pushback is telling too: a reader with attention to spare, who thinks the idea needs work, will say where and how; the depleted reader has just enough left to ask for it to be shorter.

Let’s call it manufactured incompetence, at least until a better phrase turns up. A capable person, in a role that exists for the exercise of judgement, is made unable to exercise it by the demands of the role. They then interpret that inability as a fault in the author of the idea rather than a condition of their own working life and the social structure that creates it. The environment rewards the appearance of seriousness, the full day and the fast reply, over the slower thing that seriousness is meant to protect.

It is tempting to reach for the Peter Principle¹, which describes a ceiling: you rise until you reach a job you cannot do, and there you stall. That doesn’t quite fit, I am describing a person of ample ability, in conditions that stop that ability from being expressed. If given a free week and a hard problem, that person would be formidable. The scarce resource is uninterrupted attention, not talent, and we have drifted into an arrangement in which the most senior people are the least able to access that resource.

The world is full of difficult issues, and there’s an important role for specialists in clarifying information to provide to decision makers. However there are a lot of complex issues in the world that genuinely do require deep engagement to work out a good response. The popularity of the you can fix climate change if you just plant 1 trillion trees² paper is a good example of this. The science since has pointed out that it’s much more nuanced, and that merely planting a lot of trees can be quite a bad idea in some situations, but there is a huge appeal of a simple answer.

“everything should be as simple as it can be but not simpler” Einstein³

How it is built

The people this happens to were, as a rule, excellent at something else first. They were promoted for being good at that, into a discipline nobody taught them. The Chartered Management Institute (2023) found that 82% of people entering management had received no formal management or leadership training, and 26% of senior managers and leaders had received none either.⁴ They label these as “accidental managers”. Most organisations run a single ladder: to pay someone more, or to keep them, you make them a manager, whether or not managing is something they can do or want to do.

The result is a stranding; the craft that earned the promotion moves on without them, practised now by people who still have the hours on the tools to keep up, while the craft they have now been handed was never taught to them and cannot be learned in the gaps between meetings. They sit between two competences, fluent in neither. I suspect that, in the current work climate, even if they were trained explicitly, there wouldn’t be any mention of how to engage with complex issues that can’t be thought through in a 10 slide deck.

Why the anxiety comes out as busyness

The pressure to look busy is, for most people, not vanity. It is imposed externally: a calendar filled by other people, a manager who reads a quiet hour as spare capacity, a working culture in which “you’re not paid to think” can still be said with a straight face. Performance of busyness is real, but it is the smaller part. The larger part is imposed, and the imposition rests on an old assumption, that thinking is not work, that the real work is production, and that an hour not visibly producing is an hour going spare. Newport (2024) calls this pseudo-productivity: because the output of knowledge work is hard to measure, visible activity stands in for it, as the thing that can be seen.⁵

There is a long genealogy to why we treat visible labour as virtue. Weber (1905/2001) traced the modern version to Calvinism and the doctrine of election, the belief that God has already chosen who will be saved and that no human act can move the verdict⁶. The uncertainty this produced was unbearable, so believers searched their lives for signs that they were among the elect, and steady worldly diligence became, not the cause of salvation but its evidence. Idleness, by the same logic, was damning. “Waste of time,” Weber wrote, “is the first and in principle the deadliest of sins.” The theology ebbed away and the reflex stayed. We still treat the empty hour as guilty until it proves itself productive.

Once work carries moral weight, busyness can be shown off. Bellezza, Paharia and Keinan (2017) documented the modern convention: the overbooked, never-off professional is read as important, on the inference that a person genuinely in demand would have no time to spare⁷. Veblen had described the opposite convention a century earlier, when the wealthy signalled rank by conspicuous abstention from work, his leisure class being people who did not have to earn a living at all (Veblen, 1899). As the leisure class has essentially disappeared, the professional, who does have to earn, now advertises worth through the full calendar rather than the empty one.

Traditionally the bottleneck has been caused by production. All the writing and coding and cutting and pasting has taken up weeks of effort, so for most people there’s been plenty of interstitial time for thinking about what they’re doing. As AI tools are more integrated into our work, the production is becoming a more trivial part of the process, the opportunities for thought are being squeezed. I once had a job where I drilled holes all day, and I did a lot of thinking while I was doing that. But now I am often working on two or more codebases at the same time in parallel AI chat sessions, my opportunities for accidental abstract thinking have been reduced almost to zero, and if I want to retain the benefits of that, I need to make those thinking opportunities an explicit part of my life.

The traditional assumption is that production is the expensive part, and thinking happens incidentally. As the making is handed to machines, the logic inverts. Deciding what is worth making and judging whether the made thing is any good, is the work that remains, and both are kinds of thinking. AI is a “faster builder” and it’s also a filter for low-humanity work. Being good at mental arithmetic was valorised in my childhood, but now everyone has a calculator on their phone, I haven’t heard anyone mention it for years. We are automating production while still running an attention economy built for a world where production was the bottleneck. Herbert Simon (1971) noted that a wealth of information creates a poverty of attention⁸. Attention is finite and is the input the senior roles runs on. The busyness regime spends it on being seen to spend it.⁹

The two tiers separate here. For most people the constraint is external: the time is not theirs to protect, and the demand to look occupied comes down the hierarchy. At the top it turns inward, though rarely on purpose. A senior leader could clear the calendar and think. What stops them is not a manager enforcing it but an audience they cannot ignore, a board, a market, a set of peers who read reserved capacity as underuse or as risk. To take the unhurried time the work needs, they would first have to win an argument with people who have no vocabulary for legible thinking. They do not lose that argument. They never have it, because it does not occur to them that it is theirs to start.

[alt para to the one above, eventually we’ll probably merge them into one] There’s a gradient of production making (code, drawings) to decision making as you go up the hierarchy¹⁰.

There is a more hopeful mechanism than the one above, and it does not depend on senior people ever having had spare production hours to reclaim. It needs one more piece of scaffolding first, because “senior people don’t do production” is not new information created by AI, it is the standing condition of a whole band of most hierarchies, and has been for a long time. Long before agents, team leads, department heads, directors, partners already spent their entire day on judgement-shaped work: reviewing what others made, approving or declining it, allocating scarce time and money, deciding the exception because the rule ran out. None of that is typing or drafting, and none of it goes away when AI takes over typing and drafting. It is a different, older layer from the purely information-distilling middle manager, a layer the next section argues has largely disappeared; this band makes real decisions, constantly, and has existed for as long as organisations have had more than two levels. AI eating “production” does nothing for it directly, because production was never where its day went.

The more precise version of automation as delegation, then, is not that agents free this band by taking over its typing, most of it did little typing to begin with. It is that agents start absorbing a slice of the judgement-shaped work the band already does all day: the routine review, the first-pass triage, the decision that fits an established pattern closely enough not to need a fresh mind on it. If that supervision genuinely costs less attention than supervising a person did, the freed capacity can move up, absorb some of what used to be escalated, and free the tier above in turn. Call it automation as delegation: not more hands doing the same work, but the work of deciding itself moving down a level, floor by floor, until less of it has to travel all the way to the top.¹¹ On this account manufactured incompetence at the senior tier is not solved by handing that tier more free hours directly, since it was never spending those hours on production. It is solved, if it is solved, by pushing judgement down far enough, and with enough confidence, that the top has fewer decisions left to make, only the few that are genuinely its own.

The hope carries a condition, and the essay has already named it once, a few sentences up. Delegation only frees the delegator’s attention if it is trusted enough to be left alone; oversight that still has to check everything is not delegation, it is what one recent account of AI in management calls “directed outsourcing with retained accountability,” execution handed off, judgement and liability kept.¹² If that is the more accurate description of how agents are actually being used, then pushing decisions down does not shrink anyone’s load at any layer of the band just described, it widens the span of it: more decisions get made below you, and you are still on the hook to check more of them, only faster and less visibly. That is not freed attention, it is the same attention smeared across a wider front, Leroy’s residue problem again, relocated a floor up rather than resolved. Automation-as-delegation and manufactured incompetence are both live outcomes of the same technology, and which one an organisation gets depends on whether it can extend to an agent the kind of unaudited trust it has historically struggled to extend to a junior colleague, for the same audience-shaped reasons already described. More of the day either way moves into what the next section calls manager-mode, evaluating and passing on rather than shaping. Whether that leaves the higher-order work, the judgement itself, any better resourced, or simply moves the same shortage up a floor, is the question the rest of this essay is really asking. Two things would have to be true for the hopeful version to win, and they are the subject of the next two sections: a way to tell which decisions are worth the trip to the top, and a way of getting the top enough information to tell.

Which decisions are yours

Automation as delegation, in its most hopeful reading, frees attention rather than merely relocating it. But freed attention is not self-directing. Handed back an hour, the reflex this essay has been describing does not ask what the hour is for; it looks for something to fill it with, and busyness is very good at supplying candidates. What is missing is not more free time but a way of sorting the decisions that land on a freed desk into the ones that are actually the freed person’s to sit with and the ones that were never worth the trip up the hierarchy in the first place.

Jeff Bezos gave the clearest public version of that sort in his 2015 letter to Amazon’s shareholders. Some decisions are consequential and hard, or impossible, to reverse, one-way doors: walk through, dislike the view, and there is no walking back. Most decisions are not like that; they are two-way doors, correctable at modest cost, and the right move on a two-way door is to make it quickly, on incomplete information, because the delay costs more than an occasional wrong guess.¹³ His complaint was that organisations drift, as they grow, into using the one-way-door process, slow, consultative, covering every base, on decisions that are actually two-way doors, which is a tidy technical name for exactly the busyness this essay has been describing: heavyweight seriousness spent on a decision that did not need it, at the cost of the attention the genuine one-way doors were owed.

I first heard a version of the same sort from Patrick Collison, in an interview where I understood the interviewer to be relaying a distinction they’d read rather than coining it themselves, which suggests the idea has been arrived at independently more than once, always by people who spend their day fielding more decisions than they have judgement to spare for. Collison’s version adds a second axis to Bezos’s one: not just how reversible a decision is, but how much it matters if you get it wrong. A decision that is both hard to reverse and high-stakes deserves real deliberation. One that is either easily undone or simply doesn’t matter much should be made fast, on a guess, precisely so the saved attention can go toward finding out sooner whether the guess was right.¹⁴ Put the two frameworks together and the test a freed hour needs is not “is this decision annoying” but “which door is this, and how much does it matter if I choose wrong,” a question that can be taught, unlike taste, and asked in seconds, unlike a strategy review.

Applied back to the opening scenario, the senior person skimming a difficult idea and asking for it to be made shorter is running a one-way-door problem through two-way-door process: the idea, if it is genuinely load-bearing, is exactly the kind of thing that is expensive to get wrong and hard to undo once acted on or dismissed, and it is being decided at the speed appropriate to choosing which call to skip. Manufactured incompetence and Bezos’s organisational drift turn out to be the same failure seen from opposite ends. His firms over-invest process in decisions that never needed it; my senior reader under-invests attention in the one decision that did. Both come from the same missing skill, not knowing which door you are standing in front of, and both are what happens when the sorting is never taught, never made explicit, and left instead to whatever a depleted attention defaults to, which is to treat everything as a two-way door because that is the only kind there is time left to open.

Getting the information to the decision

Knowing which door a decision is does no good if the decision never reaches you in a state you can judge. Hierarchies are, among other things, filtering devices: information about what is actually happening is generated at the bottom, closest to the problem, and by the time it has been summarised, cleared and carried up far enough to reach someone with the authority to act on it, most of what would have let them tell a one-way door from a two-way one has been filtered out along the way. This essay opened with a specialist’s job of packaging a complex idea for someone with no time to spare; that packaging is not a failure of the system, it is a feature of it, and the same feature is what strands the freed, well-triaged attention of the previous section with nothing accurate to apply itself to.

General Stanley McChrystal ran into the sharpest version of this problem commanding the Joint Special Operations Task Force in Iraq from 2003. Al-Qaeda in Iraq operated as a fast, decentralised network; the task force, however good its people, was organised as a set of stovepiped specialist units, each holding a piece of the picture and none holding the whole of it, so decisions kept climbing to the top not because they were genuinely one-way doors but because nobody lower down had enough of the picture to tell. His account of the fix, written up with Tantum Collins, David Silverman and Chris Fussell as Team of Teams, was to stop routing information up a chain of command and back down as orders, and instead push it sideways too, so that every part of the organisation had something closer to the same view of the whole as the people at the top, a condition they called shared consciousness.¹⁵ The point was not to remove hierarchy or flatten authority for its own sake; it was that decisions could only be pushed down to the people fast enough and close enough to make them well once those people had stopped being the last to know.

The structure that did the actual work of manufacturing that shared consciousness, rather than merely wishing for it, was the fusion cell: a standing arrangement that sits people from otherwise separate units, intelligence, operations, whichever specialisms the problem touches, in the same room with the same feed of information, so that what one part of the organisation learns is available to the rest of it as it happens rather than at the next scheduled report. It is Fussell’s own master’s thesis, written at the Naval Postgraduate School with Trevor Hough and Matthew Pedersen, that did the empirical work of asking what made these cells effective rather than merely present, and the answer was not the org chart but the practice: co-location, a shared information feed, and enough delegated trust that someone in the cell could act on what they were seeing instead of relaying it upward and waiting.¹⁶ The popularised version in Team of Teams is the argument; the thesis is the evidence that the argument survives contact with an actual organisation rather than staying a slogan about transparency.

None of this is a claim that a firm or a design practice should adopt military structure; it is a claim that the specific failure McChrystal was solving, decisions escalating past the people equipped to make them because information never reached the people equipped to make them, is the organisational form of manufactured incompetence, not a metaphor for it. It is also, looked at from the other side, the missing piece in the more troubling account of automation-as-delegation from earlier: “directed outsourcing with retained accountability” is exactly what you get when decisions are pushed down without the information being pushed sideways, execution moves, judgement stays put, because whoever is meant to exercise it lower down was never given enough of the picture to be trusted with it, and the person at the top keeps checking because checking is the only form of confidence the arrangement has left them. Genuine delegation, of the kind that actually frees attention rather than smearing it across a wider front, needs the same fix McChrystal needed: the people making the fast, two-way-door calls need to be looking at something close to the same information the person above them would have used, or the trust that would let those calls stand unchecked never arrives.

Two conditions, then, not one. A person needs the skill to sort what lands on their desk by which door it is behind. An organisation needs to have already solved the older, less glamorous problem of getting accurate information to wherever that sorting is happening, rather than hoarding it at the level that used to have the most spare time to read it. Together, and only together, they are what would let the freed hour from automation as delegation land on the right desk, already carrying enough to be judged, instead of arriving stripped of everything that would have told its recipient which kind of decision it was. What that hour is actually for, once it does land well, is the question the next two sections take up.

The maker who forgot they were a maker

Graham (2009) drew the useful distinction between the maker’s schedule and the manager’s¹⁷. Makers needs long unbroken stretches; a single meeting dropped in after lunch splits the day into two pieces too small to build anything in. Managers live in interchangeable hourly boxes and cannot see the problem. Graham’s own fix was to bolt his meetings to the end of the day and keep the mornings whole.

< roughly edited as far as here >

At least in the world I interact with, there are very few pure managers, the layers of management that were there purely for upward information-distillation seem to be gone. Graham sorts people into makers and managers, but realistically they’re more opposing mindsets than distinct groups.¹⁸ Everyone makes things, everyone who devises “courses of action aimed at changing existing situations into preferred ones” (Simon, 1969) is designing. It might be worth thinking about leaders as makers of choices. The mind behind a material artifact is doing the same processes that the mind behind a sales plan or a social policy is doing. A strategy document, a contract, a restructure strategy, a fee model, a decision about who sits next to whom: all of these are designed, and designing them well is challenging. The more useful cut runs not between maker-people and manager-people but between two modes the same person moves through. There is maker-mode, in which a thing is held long enough to be shaped, and manager-mode, in which inputs are received, sorted, and passed on.

What the leader has lost is not the capacity to make. It is the recognition that their own highest-value output is a made thing, with a maker’s requirements. They have been alienated from the craft nature of their own work.¹⁹ Judgement, synthesis, the ability to stay inside a hard problem until it yields: these are not the overhead around the real job, they are the job. Treated as overhead, they get the scraps of attention left once the calendar has taken its cut, and the scraps are not enough. When a difficult idea then arrives and will not go in, the natural move for someone who no longer believes thinking is work is to conclude that the idea, not the depleted attention receiving it, is at fault. The cost of switching is real and it compounds. Leroy (2009) showed that when we move between tasks a residue of the last one lingers and drags on the next, so a day chopped into fragments is not a day of many small productivities but a day spent mostly in the smear between them.²⁰ [this para feels like we’re getting a bit lost. We’re talking too much about being a maker, and not enough about being a judge. The maker angle is great, and I think we should keep it but it’s disappearing down its own rabbit hole. the core of this argument is really about attention, and having what you need to be able to bring the force of that attention to bear on a problem. The following section on leisure needs to stay attached to that core idea, so it doesn’t just get too excited about leisure as a good in itself, or distracted down a deep-work rabbit hole, etc.]

What leisure was for

The Greek is scholē (the origin of our “school”)²¹ did not mean rest or amusement, it meant freedom from the pressure of necessity, the condition under which the highest activity, contemplation, became possible. Aristotle’s (ca. 350 B.C.E./2009) formula is that we are busy in order to be at leisure, not the reverse; work is for the sake of scholē, and to invert that is to mistake the means for the end. One concession before leaning further on him: the leisure Aristotle prized was for contemplation pursued as an end in itself, not for getting through a strategy paper faster. I borrow scholē for its shape, freedom from the pressure of necessity, and for its etymology, not to smuggle in a claim that leaders need philosophy. It is the shape the modern worker has lost, and it will hold deep reading, incubation or hard judgement equally well. The Latin word for leisure was otium, and their word for business, for the state of not being at leisure, was negotium, which is simply nec-otium, the negation of leisure. Business was defined, in the language itself, as the absence of the condition in which thought happens.

One thing to bear in mind is that Athenian leisure rested on people who had none, on enslaved workers and on women whose labour was not counted. Scholē was an unacknowledged class privilege before it was a philosophical ideal, and any argument that reaches for it has to engage with that.²² The Greeks were right about what leisure was for even while being wrong about who deserved it. The uncomfortable question is not whether contemplative freedom is worth having but who gets to have it, and at whose expense.

Pieper’s (1948/1998) argument in Leisure: The Basis of Culture is that leisure is not the reward that follows work but the precondition of any work worth doing, a stance of receptive openness in which real insight can arrive, and that a culture of “total work,” which grants legitimacy to nothing that is not visibly labour, corrodes the very faculty that thought depends on.²³ Set against the leader who cannot afford to be seen doing nothing, Pieper is not quaint. He is describing the exact organ that manufactured incompetence removes. [I don’t understand this well enough, so I don’t think the reader will understand it either.]

The disciplines that kept a theory of this survived better. Mathematics and physics prize hard grinding, and they also honour the person lying on a sofa staring at the ceiling, because they have never confused the second with idleness. Wallas (1926), a co-founder of the London School of Economics, set out the stages of creative thought a century ago, and the load-bearing one is incubation, the interval in which you deliberately stop working on a problem so the mind can work on it without you.²⁴ He drew it partly from Poincaré’s (1908/1952) account of a solution to an abandoned problem arriving as he stepped onto a bus with his mind entirely elsewhere. These fields have a doctrine that sanctions the empty afternoon. They can point to it and say that this is a stage of the work, not a lapse from it.

The creative professions have no such doctrine, and the reason is legibility rather than temperament. A proof either checks or it does not, so mathematics can be indifferent to how you reached it; the result vouches for the process after the fact. Design output cannot vouch for itself the same way. It is satisficing, in Simon’s sense, a good-enough choice pulled from a solution space far too large to search to the end, and nobody can show that this plan is the best plan, because the search that would prove it never terminates. Into that gap steps process as evidence: unable to prove the output is good, you prove instead that you worked hard for it, and hard work means visible hours. Architecture then adds a myth of its own, the heroic gesture, the master’s sketch thrown down whole on a napkin, which worships the instantaneous flash and so credits neither the long grind nor the slow incubation. The field ends with the worst of both cultures. It will not count thinking as work, and it will not count patience as thinking.

What it costs, and to whom

There is an economics to this that is more than metaphor. Cyert and March named organisational slack decades ago and argued that it is not waste but reserve, the spare resource that lets a firm absorb a shock or pursue something new instead of merely surviving the day (Cyert & March, 1963).²⁵ DeMarco put it more sharply, borrowing from queueing theory: as any system nears full utilisation its ability to respond does not fall away gently, it collapses, because a queue at a hundred percent has nowhere to put the next arrival (DeMarco, 2001).²⁶ An organisation that runs its people at total capacity has optimised away the very thing that lets it think and adapt, and has done so while congratulating itself on efficiency.

The name for the loss is X-inefficiency. Leibenstein coined it for firms that produce less than their inputs allow, not by misallocating resources in the textbook sense but through slack inside the organisation, effort withheld, attention misspent (Leibenstein, 1966).²⁷ Usually the word slack is an accusation, and there is a tension worth naming rather than smoothing over: Leibenstein’s slack is the disease, and I am about to call a kind of slack the cure. They are two different things wearing one word. Idle effort is waste. Reserved attention, held so that thought and response stay possible, is not. A regime that stamps out every visible gap removes both, and in removing the second it lowers the output it set out to raise. Running people at full tilt does not maximise what they produce. It maximises what they produce that can be seen.

That the loss is real and measurable is no longer arguable. The World Management Survey, run by Bloom and Van Reenen across tens of thousands of firms, found that management practice is strongly tied to productivity, profitability and survival, that badly managed firms form a long persistent tail, and, in the largest study, that management practices account for more than a fifth of the variation in productivity between plants, a share comparable to or larger than research spending, technology or workforce skill (Bloom & Van Reenen, 2007, 2010; Bloom et al., 2019).²⁸ Management is not the soft residue left over once the real inputs are counted. It is one of the real inputs, and how a leader spends their attention is much of what the word means.

If firms routinely produce below their potential, then society gets less than it could from the same people, materials and time, and the shortfall is real output that never came into being. I am inclined to go one street further and say that because the state supplies so many of the inputs, the educated people, the infrastructure, the stable institutions, a chronically under-producing private sector returns less to the state than it received, which shows up as a tax burden heavier than it needed to be and private exchanges thinner than they might have been, a slow leak in the collective account.²⁹ That last step is mine, and I hold it more loosely than the rest; the economics I can cite stops at the firm. The firm-level claim alone is enough to make the point uncomfortable. The leader who cannot find an hour to think is not indulging a private vice. They are a small recurring loss in a very large ledger.

We (architects) do not, for the most part, bill clients by the hour. We agree a fee for a phase of work and live with the difference between that fee and what the work actually costs us; the timesheets we keep are internal. On paper this is outcome pricing, and it ought to free us from the tyranny of the billable minute. It does not, quite, and I have started to suspect why. Under a fixed fee, an hour spent thinking rather than producing is not merely unbillable, which would make it invisible. It is worse than invisible, because it is a direct subtraction from the margin on a pot that cannot grow. The rational response to a capped fee is to compress the hours, and the hours easiest to compress are the quiet ones with nothing to show. So the instrument meant to buy us freedom to think may be the one most efficiently taxing the thinking away. I do not know that this is true. I know it is the shape of the thing I feel.

There is a version of this that ends well. We are building machines to do the producing, and if the old excuse for filling every hour was that the hour could be spent making something, that excuse is dissolving. The freed capacity could go to the thing the making was always meant to serve, deciding what is worth doing and thinking hard about whether we are doing it. It could equally be swallowed whole and poured back into doing more, faster, because the reflex that reads an empty hour as waste will not retire simply because the work has changed. The Greeks bought their leisure with other people’s unfreedom. We may be about to buy ours with machines, and the open question is whether we will have kept the faculty to use it, or whether we will only have found new hours to fill. This is the first time in history that humanity has the chance to carve out ethical scholē—where the “unfree labour” keeping the system running is on a non-biological substrate.³⁰

We have built work that depends, at its most valuable, on unhurried attention, and then arranged our institutions so that unhurried attention is the one thing a serious person may not be seen to take. The senior person skimming the difficult idea is not failing at the job. They are succeeding at the job we actually set, which is to appear fully occupied at all times, and that success is what makes the other job, the real one, impossible. Negotium was the negation of leisure. We have built a whole economy of it and called the negation the work.

References

Aristotle. (2009). The Nicomachean ethics (W. D. Ross, Trans.; L. Brown, Ed.). Oxford University Press. (Original work published ca. 350 B.C.E.)

Baird, B., Smallwood, J., Mrazek, M. D., Kam, J. W. Y., Franklin, M. S., & Schooler, J. W. (2012). Inspired by distraction: Mind wandering facilitates creative incubation. Psychological Science, 23(10), 1117–1122. https://doi.org/10.1177/0956797612446024

Beaty, R. E., Benedek, M., Wilkins, R. W., Jauk, E., Fink, A., Silvia, P. J., Hodges, D. A., Koschutnig, K., & Neubauer, A. C. (2014). Creativity and the default network: A functional connectivity analysis of the creative brain at rest. Neuropsychologia, 64, 92–98. https://doi.org/10.1016/j.neuropsychologia.2014.09.019

Bezos, J. (2015, April 6). 2015 letter to shareholders. Amazon.com. https://s2.q4cdn.com/299287126/files/doc_financials/annual/2015-Letter-to-Shareholders.PDF

Bellezza, S., Paharia, N., & Keinan, A. (2017). Conspicuous consumption of time: When busyness and lack of leisure time become a status symbol. Journal of Consumer Research, 44(1), 118–138. Conspicuous Consumption of Time: When Busyness and Lack of Leisure Time Become a Status Symbol

Bloom, N., Brynjolfsson, E., Foster, L., Jarmin, R., Patnaik, M., Saporta-Eksten, I., & Van Reenen, J. (2019). What drives differences in management practices? American Economic Review, 109(5), 1648–1683. What Drives Differences in Management Practices?

Bloom, N., & Van Reenen, J. (2007). Measuring and explaining management practices across firms and countries. The Quarterly Journal of Economics, 122(4), 1351–1408. Measuring and Explaining Management Practices Across Firms and Countries*

Bloom, N., & Van Reenen, J. (2010). Why do management practices differ across firms and countries? Journal of Economic Perspectives, 24(1), 203–224. Why Do Management Practices Differ across Firms and Countries?

Chartered Management Institute. (2023). Taking responsibility: Why UK plc needs better managers. Chartered Management Institute.

Collison, P. (2018, December 24). The Tim Ferriss Show (No. 353) [Audio podcast episode]. Tim Ferriss. https://tim.blog/2018/12/24/the-tim-ferriss-show-patrick-collison/

Cyert, R. M., & March, J. G. (1963). A behavioral theory of the firm. Prentice-Hall.

DeMarco, T. (2001). Slack: Getting past burnout, busywork, and the myth of total efficiency. Broadway Books.

Fussell, C. L., Hough, T. W., & Pedersen, M. D. (2009). What makes fusion cells effective? [Master’s thesis, Naval Postgraduate School]. Calhoun: NPS Institutional Archive. https://calhoun.nps.edu/handle/10945/4438

Graham, P. (2009, July). Maker’s schedule, manager’s schedule. Maker’s Schedule, Manager’s Schedule

Leibenstein, H. (1966). Allocative efficiency vs. “X-efficiency.” The American Economic Review, 56(3), 392–415.

Leroy, S. (2009). Why is it so hard to do my work? The challenge of attention residue when switching between work tasks. Organizational Behavior and Human Decision Processes, 109(2), 168–181. https://doi.org/10.1016/j.obhdp.2009.04.002

McChrystal, S., Collins, T., Silverman, D., & Fussell, C. (2015). Team of teams: New rules of engagement for a complex world. Portfolio/Penguin.

Newport, C. (2024). Slow productivity: The lost art of accomplishment without burnout. Portfolio/Penguin.

Oppezzo, M., & Schwartz, D. L. (2014). Give your ideas some legs: The positive effect of walking on creative thinking. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40(4), 1142–1152. APA PsycNet

Peter, L. J., & Hull, R. (1969). The Peter principle: Why things always go wrong. William Morrow.

Pieper, J. (1998). Leisure: The basis of culture (G. Malsbary, Trans.). St. Augustine’s Press. (Original work published 1948)

Poincaré, H. (1952). Mathematical creation. In B. Ghiselin (Ed.), The creative process: A symposium (pp. 22–31). University of California Press. (Original work published 1908)

Simon, H. A. (1969). The sciences of the artificial. MIT Press.

Simon, H. A. (1971). Designing organizations for an information-rich world. In M. Greenberger (Ed.), Computers, communications, and the public interest (pp. 37–72). Johns Hopkins Press.

Sio, U. N., & Ormerod, T. C. (2009). Does incubation enhance problem solving? A meta-analytic review. Psychological Bulletin, 135(1), 94–120. APA PsycNet

Veblen, T. (1899). The theory of the leisure class: An economic study of institutions. Macmillan.

Wallas, G. (1926). The art of thought. Jonathan Cape.

Weber, M. (2001). The Protestant ethic and the spirit of capitalism (T. Parsons, Trans.). Routledge. (Original work published 1905)

Appendix: imagined critiques

A panel of readers, each responding to a specific claim above rather than the essay in general. Sorted alphabetically by surname.

Hannah Arendt

You are running two different things together under the word “judgement,” and the conflation is costing you. There is work — fabrication, the maker shaping a durable thing, which is what your Graham section and your maker-mode/manager-mode distinction actually describe. And there is action — plural, unpredictable, exercised among others, disclosed in speech, which is what a senior person exercising judgement on a contested idea is actually doing. You treat both as versions of the same attention-starved production problem, solvable by the same fix (protect the hours, sort the inputs). But action doesn’t fabricate an output the way work does, and Bezos’s door-sorting, however useful for two-way operational calls, is a fabrication-logic tool. Applied to genuine deliberation among peers over a hard idea, it risks producing exactly the “shorter, please” reflex you’re diagnosing — efficient handling of what should have been left open.

Your closing hope — that AI-freed hours go to “actual judgement” — still describes vita activa reabsorbing vita contemplativa into productivity. Contemplation, for Aristotle as much as for me, is not supposed to produce a decision faster. If the freed hour is only justified because it yields better output, you haven’t recovered scholē, you’ve optimized total work more cleverly. That is the total-work logic Pieper is trying to name, and your own resolution may be quietly inside it.

Peter Drucker

You’ve rediscovered The Effective Executive and not cited it, which surprises me more than anything else in the piece — time as the one truly scarce, non-substitutable resource; the need to consolidate discretionary time into large blocks rather than let it be shredded; the discipline of knowing where your time actually goes rather than where you think it goes. Your “which decisions are yours” section is good but incomplete without the companion discipline: not just sorting by reversibility and stakes, but asking “what, of everything on my desk, only I can do” — delegation isn’t just trust in the delegate, it’s rigor in the delegator about what cannot be delegated at all.

Where this essay is diagnosis-rich and prescription-poor is exactly where I’d press it. Naming manufactured incompetence is useful; it does not yet tell a reader what to do Monday morning. Effectiveness is a discipline, practiced, not a condition restored by having more free hours appear. Give AI back the hours it frees and nothing changes unless the executive has already built the habit of time-logging and boundary-setting beforehand — otherwise the freed time is simply recolonized by the next urgent thing, which your own essay half-admits in the closing line about “new hours to fill.” I’d want a closing move from diagnosis to habit.

Benedict Evans

The interesting question isn’t “will AI free attention” — it’s “does freed capacity ever actually get spent on the thing it was freed for, or does something else expand to fill it.” History’s answer for almost every productivity tool is the latter: PCs and email were supposed to free management time and instead generated more meetings, more CC lists, more expected responsiveness. Your own closing line half-concedes this (“we may only have found new hours to fill”) but then the essay still frames AI as offering “the first chance at ethical scholē,” which is a much bigger claim than the evidence in the piece supports. I’d want that softened or tested against the base rate: every prior wave of automation freed some layer’s time and the org absorbed it into more oversight, more coordination, more reporting — this is just Jevons’ paradox applied to management attention.

“Directed outsourcing with retained accountability” is the realistic default, not the disappointing exception — that’s what happens whenever a new capability arrives faster than the trust and legal/liability structures needed to actually delegate to it. I’d anchor the piece less in what AI could free up in principle and more in what specifically has to change in accountability structures before any of this delegation becomes real, because right now the essay’s hope rests entirely on an “if” it never resolves.

Silvia Federici

Footnote 12 is the most honest moment in the essay and the most quickly abandoned. You concede Athenian scholē rested on enslaved and uncounted women’s labour, then move straight to “we may be about to buy ours with machines” — as if the accounting problem were a historical curiosity rather than the live question. Whose labour is running the rest of a contemporary senior person’s life so that they can have protected attention? An assistant’s calendar triage, a spouse’s uncounted household and emotional management, a nanny, a cleaner — these are today’s version of the domestic labour Athens didn’t count either. Protected attention has never been free; it has always been subsidized by someone whose time doesn’t appear in anyone’s ledger. Your essay’s economics section (Bloom, DeMarco, Leibenstein) counts firm-level slack and productivity variance but never asks whose unpaid hours make a firm’s visible slack possible at all.

The same blind spot follows your AI move. You raise the moral-status question of the AI itself in the bio footnote, admirably, but skip the question I’d ask first: whose labour built and maintains the thing — the data annotators, the content moderators, the people mining the inputs, the water and power drawn from wherever the datacenter sits. “Ethical scholē” is not available to declare in advance of that inventory. The free lunch always has a kitchen; find out who’s in it before calling the meal ethical.

David Graeber

Handsome essay, wrong villain. You treat busyness as an emergent tragedy — nobody’s fault, a reflex left over from Calvin, a coordination failure everyone would fix if they could see it clearly. I’d ask: whose interest does it serve that senior judgement stays scarce and performative? Bullshit Jobs argues managerial feudalism runs on loyalty rituals, not output — the full calendar is tribute, not accident. Your “accidental managers” are, in my terms, doing a job that exists to demonstrate fealty up the chain, and the CMI stat about 46% believing promotions go on relationships rather than ability supports that reading better than the training-gap reading you gave it.

The scholē section is the most honest thing here — crediting that Athenian leisure was slave-built is the right move, but then the essay pivots straight to “AI frees us the same way, ethically this time,” which is exactly the Keynes 1930-fifteen-hour-week mistake: productivity gains get captured by whoever owns the surplus, not distributed as leisure to whoever generated it, unless power relations change. Nothing in this essay’s fixes (sort decisions better, route information better) touches ownership or power. They’re managerial-technocratic fixes to what I’d call a political problem. Worth at least naming that the “hopeful” ending requires a distributional fight the essay doesn’t stage.

Melissa Gregg

I’d place your essay inside the genre it’s trying to diagnose. Time-management literature has been telling this exact story since the 1950s: a new technology or technique (the day planner, the Filofax, GTD, now AI) promises to finally free the knowledge worker’s attention, the failure to realize that promise gets individualized as a skill the worker lacks — sorting decisions well, in your version — and the structural conditions producing the overload go unexamined for another cycle. Your “two conditions” (sorting skill, information flow) are elegant, but they are still a personal-and-organizational-competence fix to what your own cost section (Bloom & Van Reenen, DeMarco) frames as a structural, firm-level problem. Those two registers don’t fully meet.

I’d also ask why the fix stops at skill and information rather than headcount, workload, or the fee structures you flag honestly in the architecture aside. That aside is the most materially concrete thing in the piece — a fixed fee that taxes quiet hours directly — and it points toward a structural rather than a dispositional cause. Time-management writing has a long history of stopping one step short of that admission, because the skill-gap story is more sellable and less threatening to who holds power over workload than the fee-structure or staffing story is. I’d want the essay to take its own best example more seriously than its own resolution does.

Dan Hill

I like this a lot more at the systems-design level than the essay currently operates. McChrystal’s fusion cells are the right instinct — co-location, shared feed, delegated authority — but that’s an org-design and spatial intervention, and the essay treats it as a data point rather than a template to build from. Same with the architecture fixed-fee aside near the end — that’s not a footnote, that’s the whole argument in miniature: procurement models silently determine whether thinking time can exist at all, and you’ve buried your best, most concrete, most “dark matter” example under the leisure discussion instead of leading with it.

My pushback: this essay diagnoses beautifully and then hopes. Where’s the redesigned artifact — the calendar, the fee structure, the org chart, the room layout — that would actually produce protected attention rather than just naming the absence of it? “Two conditions, then, not one” (sorting skill + information flow) is a lovely, clean synthesis, but it’s still a description of what a good system would have, not a design for one. I’d want a third move: given AI’s changing what’s cheap to build, what’s the actual redesigned operating model — governance, incentives, physical or digital space — that makes manufactured incompetence structurally impossible rather than personally avoidable. Diagnosis without a proposed artifact is half a strategic design brief.

Arlie Hochschild

You state that busyness “is imposed externally… not vanity” and treat performance as the smaller part. My fieldwork for The Time Bind found something messier: offered real options to work less, many professionals didn’t take them, because the office had become a refuge from a second shift at home that was, if anything, more demanding and less rewarded. People co-produce the time bind with their employers; it is not simply done to them. I’d want to know which of your senior people, given a genuinely free week, would actually spend it on the hard idea rather than filling it with something else that felt more urgent — your own closing line half-suspects the answer, but the body of the essay is written as though the imposition were purely one-directional.

There’s a second gap: I’d bring The Managed Heart to your “accidental managers.” A person stranded between competences, doing a job they were never taught, has an emotional-labour problem, not only an attention problem — visible busyness is often a performance calibrated to manage the anxiety of being unqualified for the role you’re now in, a compensatory display rather than a pure by-product of an overfull calendar. If that’s even partly right, better information routing and better decision-sorting (your S1, S2) won’t touch it, because the busyness was never only instrumental. It was also armour.

Ursula K. Le Guin

You’ll like that I clock the “heroic gesture” line about architecture’s napkin-sketch myth near the end — that’s my territory. In The Carrier Bag Theory of Fiction I argued Western narrative is spear-shaped: a hero, a conquest, a line driven at a target. Your fixes — sort the decision, route the information, act — are spear-shaped too, even when the target is “protect attention.” A carrier-bag organization wouldn’t sort and dispatch decisions faster; it would hold them, unresolved, the way a gatherer holds a mixed handful without forcing it into a category yet. Your essay wants the second thing but keeps reaching for tools built for the first.

I’d also push on the hierarchy itself, not just what’s done to attention within it. In The Dispossessed the Odonians rotate the ugly, tedious jobs precisely so no one is stranded doing only the thing they were promoted into and nothing else — your “accidental managers,” stranded between competences, are a symptom of a single ladder, and no amount of better sorting fixes a ladder. You know this — the essay gestures at it and backs away. I’d want you to follow the gesture instead of correcting course toward the technocratic fix. The problem may not be that busyness prevents judgement; it may be that hierarchy manufactures both the busyness and the shape of the judgement it eventually allows.

Iris Murdoch

Your opening scenario deserves a sharper diagnosis than depletion alone gives it. In The Sovereignty of Good I described moral attention as “a just and loving gaze directed upon an individual reality” — the discipline of seeing a particular thing clearly instead of through the fog of a self-serving fantasy. Your senior reader who skims and pronounces the idea muddled may not only have run out of attention; they may be actively protected from looking closely by an ego that has a lot riding on staying decisive and in-demand. A genuinely difficult idea threatens that self-image in a way a familiar one doesn’t, and “make it shorter” is often not exhaustion talking but vanity — a small, comfortable evasion of the harder, more humbling act of really looking.

That reframes manufactured incompetence as partly a virtue problem, not only an infrastructure problem. Bezos’s door-sorting and McChrystal’s information-routing both optimize where attention goes; neither touches its quality once it arrives. I’d want a section on what actually cultivates the capacity for that loving, accurate gaze — habituation, practice, a kind of unselfing — sitting alongside your structural fixes, because a well-routed decision handed to an ego still defending itself will get the same shortening reflex, just faster and with better information behind it. Weil gives you attention as grace; I’d give you attention as a moral skill you can fail at even when perfectly rested.

Jenny Odell

How to Do Nothing resists exactly the move your essay makes at the end. Your own Pieper paragraph concedes that contemplation isn’t for “getting through a strategy paper faster” — and then the resolution immediately reneges on that concession, proposing that freed capacity should go to “actual judgement,” which is still attention justified by its yield. I’d ask what it would mean to defend attention without that alibi: not because protecting it produces better decisions or a fifth of a firm’s productivity variance, but because a person is owed some hours that answer to nothing and convert into no deliverable. Your economics section (Simon, Leroy, X-inefficiency) is honest about treating attention as a resource to allocate efficiently — but that framing is the attention economy’s own vocabulary, just applied more sympathetically. It still prices attention by what it’s worth to a firm.

Footnote 12 admits Athenian scholē was subsidized by unfree labour; I’d push further than admission toward refusal — don’t let AI’s freed hours inherit the same debt by immediately putting them back to work “on the thing the making was always meant to serve.” Some of them should serve nothing. That’s the whole point of leisure Aristotle actually meant, before you talk yourself out of it in the very next clause.

Elinor Ostrom

Both your fixes — Bezos’s door-sorting, McChrystal’s fusion cells — optimize inside a hierarchy without asking whether the hierarchy is the actual design flaw. Governing the Commons documents groups that manage shared, contested resources without a top imposing rules on a bottom at all: locally devised rules, monitoring by the people actually there rather than an external authority, graduated sanctions, and — critically — the right to modify the rules themselves rather than have a taxonomy (one-way door, two-way door) handed down from above and applied uniformly. Your S1 assumes decisions get sorted by a scheme designed at the top and pushed outward; my design principles assume the people closest to a decision devise their own sorting, because they hold the local knowledge that never survives the trip upward anyway — which is precisely the filtering problem your own McChrystal section names.

So “automation as delegation” as you’ve written it still means: decisions move down, but the authority to define what counts as a two-way door stays up. I’d ask whether AI actually enables something closer to self-governing units at the edge that no longer need to ask permission at all — not faster delegation within the existing ladder, but fewer rungs on the ladder mattering. That’s a structurally different bet than the one your essay makes, and historically it’s the one that produces the resilience gains, not the throughput gains.

Venkatesh Rao

This is Gervais Principle territory and doesn’t quite know it. The Clueless middle manager already is your accidental manager — I’d want the essay to engage the power structure, not just the psychology. “Manufactured incompetence” reads as something that happens to senior people, but who benefits from it? Busyness-as-legibility-theater is a control mechanism: the Sociopaths at the top don’t need to protect their own attention because they’ve offloaded judgement onto process and the Losers/Clueless absorb the cost of appearing occupied. The essay’s framing keeps this accidental and structural-but-blameless; I’d push it toward “cui bono.” Also — the AI-as-delegation section is basically asking whether AI increases or decreases legibility, and that’s the more interesting question than attention per se. An organization doesn’t fear illegible judgement because it’s slow, it fears it because it’s unauditable. “Directed outsourcing with retained accountability” isn’t a failure to trust the agent — it’s the hyperlegibility reflex working exactly as designed, protecting the org from blame, not optimizing anyone’s day. I’d retitle the resolution section around legibility vs. judgement rather than attention vs. busyness — attention is the resource, legibility is why it gets spent this way.

Scott Alexander (Siskind)

Structurally this is very ACX-compatible — hedged claims, footnoted uncertainty, an honest “this step is mine and I hold it loosely” (n.19), a caveat that incubation effects are “genuine but small” rather than oversold. Good epistemic hygiene, genuinely. My question is falsifiability: how would we distinguish “senior person’s judgement is being suppressed by busyness” from “senior person’s judgement was never that good, and busyness is just the excuse,” or from reverse causation — people who thrive on visible-activity signalling self-select into and get promoted within roles that reward it, so the depletion isn’t imposed on ability, it’s what the selection process was optimizing for all along? That’s uncomfortably close to the Peter Principle you’re at pains to distinguish, and I don’t think the contrast fully survives contact with selection effects.

I’d want a testable prediction: give a matched sample of senior reviewers either a protected unhurried block or a normal fragmented day, and see if idea-acceptance/critique-quality differs in a predictable direction. If it doesn’t, the “condition not fault” framing weakens considerably. Otherwise this risks being an appealing-because-flattering-to-the-competent story (nice things happen to smart depleted people, not incompetent ones) rather than a falsifiable mechanism. Still, genuinely enjoyable, well-sourced essay — I’d read the full version.

Simone Weil

Your Pieper paragraph is the one place you admit not understanding your own point, and I think I can tell you why: you and Pieper are describing “receptive openness,” but the essay keeps translating attention into an economic register — Simon’s scarce resource, an input to allocate — and receptive openness cannot survive that translation. Attention, properly understood, is not effort applied harder or longer. It is a suspension of the self, a waiting, empty of the will that wants a result. “Attention is the rarest and purest form of generosity.” The busyness regime doesn’t just use up a finite quantity of this thing — it trains people out of the capacity to attend at all, the way unused muscle atrophies, which is a different and worse injury than mere depletion. A freed hour handed to someone who has lost the discipline of waiting will not become judgement. It will become anxiety looking for something to fill it, which is precisely the failure mode your own final section predicts and then doesn’t quite explain.

This is also why your economic framing of attention (Simon, X-inefficiency, slack) sits oddly against your spiritual framing (scholē, acedia). They are not the same claim wearing different vocabulary, as you did with Cyert and Leibenstein. One treats attention as budget; the other treats it as a faculty that must be practiced to still exist. I would keep the second and be suspicious of how far the first can carry your argument.

Shoshana Zuboff

My early fieldwork on office automation found two possible outcomes from the same technology, “informating” or automating away judgement, and which one an organization got was determined by managerial intent, not by anything inherent in the machine. Your “automation as delegation” is the informating case, argued in hope; “directed outsourcing with retained accountability” is the automating-out-judgement case, argued as the disappointing default. I’d push you to treat the second not as a regrettable stopgap on the way to the first, but as the likely profit-maximizing design point for whoever builds and sells the agent. A platform that lets a manager retain full accountability while checking less carefully is a more valuable product to that manager’s employer than one that requires real, unaudited trust — it produces the appearance of oversight at lower cost, which is precisely what a market for managerial tools will select for.

My later work on surveillance capitalism gives a sharper version of your McChrystal material too: shared information doesn’t only enable faster judgement lower down, it also enables finer-grained extraction and monitoring of whoever is now visible in that shared feed. Fusion cells and “directed outsourcing” may be the same infrastructure serving two incompatible purposes depending on who controls it. Before calling AI a chance at ethical scholē, I’d want to know who owns the feed, because that answer, historically, has told me more than the technology’s stated intent ever has.

---

Argument map

flowchart TD
    classDef core     fill:#1e40af,color:#fff,stroke:#1e3a8a,stroke-width:2px
    classDef claim    fill:#6d28d9,color:#fff,stroke:#5b21b6,stroke-width:2px
    classDef fix      fill:#065f46,color:#fff,stroke:#064e3b,stroke-width:2px
    classDef sq       fill:#6b7280,color:#fff,stroke:#4b5563,stroke-width:1px,font-size:12px
    classDef contrast fill:#92400e,color:#fff,stroke:#78350f,stroke-width:1px,font-size:12px

    %% ── CORE SPINE ──────────────────────────────────
    P1["Complex problems genuinely require\ndeep, sustained attention to engage well"]:::core
    P2["Senior roles exist to exercise\nthis kind of judgement"]:::core
    P3["The busyness regime consumes\nall attention available for judgement"]:::core
    IC["Senior people cannot engage\nwith complex ideas presented to them"]:::claim
    MC["They misattribute this to idea quality,\nnot their own depleted condition\n═══════════════════════\nManufactured Incompetence"]:::claim
    S1["Fix requires: skill to sort\nwhich decisions need deep attention"]:::fix
    S2["Fix requires: accurate information\nreaching whoever is sorting"]:::fix
    RES["AI offers the first chance at ethical scholē\n— freed capacity could go to actual judgement —\nbut only if we defend attention\nrather than fill the new hours"]:::fix

    P1 & P2 & P3 --> IC --> MC --> S1 & S2 --> RES

    %% ── WHY BUSYNESS HAS MORAL WEIGHT (→ P3) ───────
    subgraph SG_BUSY ["Why busyness has moral weight"]
        direction TB
        SQ_CM["'Accidental managers'\n82% receive no management training\n(CMI 2023)"]:::sq
        SQ_WB["Weber: Calvinist doctrine gave\nvisible labour its moral charge;\nthe theology left, the reflex stayed"]:::sq
        SQ_BZ["Bellezza et al.: busyness now\nsignals status (inverts Veblen's\nleisure class)"]:::sq
        SQ_EX["External imposition: calendar filled\nby others; quiet = 'spare capacity'\nin others' eyes"]:::sq
        SQ_NP["Newport: pseudo-productivity —\nvisible activity stands in for\nknowledge-work output"]:::sq
        SQ_CM ~~~ SQ_WB ~~~ SQ_BZ ~~~ SQ_EX ~~~ SQ_NP
    end
    SG_BUSY --> P3

    %% ── ATTENTION AS THE SCARCE RESOURCE (→ P3) ────
    subgraph SG_ATT ["Attention as scarce resource"]
        direction TB
        SQ_SI["Simon (1971): wealth of information\ncreates a poverty of attention"]:::sq
        SQ_AI["AI compresses production;\nremoves interstitial thinking time;\nattention becomes the bottleneck"]:::sq
        SQ_LR["Leroy: attention residue —\nfragmented days compound the\ncost of every context switch"]:::sq
        SQ_SI ~~~ SQ_AI ~~~ SQ_LR
    end
    SG_ATT --> P3

    %% ── CONTRAST: DIFFERENT FAILURE MODE (→ MC) ────
    SQ_PP["Peter Principle: different failure —\ngenuine inability, not prevented ability.\nFix is better promotion decisions.\nHere, better decisions wouldn't help."]:::contrast
    SQ_PP -.->|"contrast"| MC

    %% ── SORTING DECISIONS (→ S1) ────────────────────
    subgraph SG_SORT ["Sorting decisions"]
        direction TB
        SQ_BE["Bezos: one-way vs two-way doors\n(reversibility × stakes)\nMost decisions are two-way;\ntreating them as one-way is the drift"]:::sq
        SQ_CO["Collison: same sorting framework,\narrived at independently — suggests\nit's discovered by need, not theory"]:::sq
        SQ_BE ~~~ SQ_CO
    end
    SG_SORT --> S1

    %% ── INFORMATION FLOW (→ S2) ─────────────────────
    subgraph SG_INFO ["Information flow"]
        direction TB
        SQ_MC["McChrystal: hierarchies filter info\nas it rises — JSOC couldn't act fast\nbecause nobody lower had the whole picture"]:::sq
        SQ_FC["Fusion cells: co-location +\nshared information feed +\ndelegated authority to act"]:::sq
        SQ_MC --> SQ_FC
    end
    SG_INFO --> S2

    %% ── AUTOMATION AS DELEGATION (→ RES) ────────────
    subgraph SG_DEL ["Automation as delegation"]
        direction TB
        SQ_DP["Delegation = Programming (2015):\na function call indifferent to what\nsits on the other end"]:::sq
        SQ_AD["AI as delegation: pushes judgement\ndown the hierarchy, freeing\nsenior attention — in principle"]:::sq
        SQ_DA["Condition: needs real trust.\n'Directed outsourcing with retained\naccountability' just smears\nthe same attention wider"]:::sq
        SQ_DP --> SQ_AD --> SQ_DA
    end
    SG_DEL --> RES

    %% ── WHAT PROTECTED ATTENTION LOOKS LIKE (→ RES) ─
    subgraph SG_PROT ["What protected attention looks like"]
        direction TB
        SQ_GR["Graham: maker-mode needs\nunbroken stretches; a single\nmeeting splits the day in two"]:::sq
        SQ_SC["Aristotle: scholē — freedom from\nnecessity is the precondition\nof judgement (negotium = nec-otium);\nnote: built on unfree labour"]:::sq
        SQ_PI["Pieper: 'total work' corrodes\nthe very faculty that thought\ndepends on (acedia as hidden busyness)"]:::sq
        SQ_WL["Wallas / Poincaré:\nincubation is a legitimate stage\nof work, not a lapse from it"]:::sq
        SQ_LD["Maths & physics have a doctrine\nthat sanctions the sofa;\ndesign and architecture do not"]:::sq
        SQ_GR ~~~ SQ_SC ~~~ SQ_PI ~~~ SQ_WL ~~~ SQ_LD
    end
    SG_PROT --> RES

    %% ── WHY THE FAILURE IS COSTLY (→ MC) ────────────
    subgraph SG_COST ["Why the failure is costly"]
        direction TB
        SQ_BV["Bloom & Van Reenen:\nmanagement quality explains 20%+\nof productivity variance across firms"]:::sq
        SQ_LB["DeMarco / Leibenstein:\nslack is productive reserve, not waste;\nat full utilisation, responsiveness collapses"]:::sq
        SQ_AF["Architecture's fixed fee:\nquiet hours cost margin directly —\nthe instrument meant to free thinking\nmay tax it most efficiently"]:::sq
        SQ_BV ~~~ SQ_LB ~~~ SQ_AF
    end
    SG_COST --> MC

---

Reading the map

Core spine (blue/purple → green): The argument runs in a straight chain — complex problems need attention; senior roles exist to provide it; busyness consumes it; therefore the judgement fails; the failure is misread as the idea’s problem; fixing it requires two simultaneous conditions; AI offers a rare opening.

Side quests (grey): Each supports a specific node on the spine but doesn’t change the logical chain if removed. They answer why busyness has its grip (Weber, Bellezza, CMI), why attention is the real scarcity (Simon, Leroy), how you would actually implement the two fixes (Bezos, McChrystal), and what protecting attention looks like in practice (Graham, Aristotle, Wallas).

Contrast node (brown, dashed): The Peter Principle is deliberately distinguished — it’s a different diagnosis, included to sharpen rather than support the main claim.