A Poverty of Attention

The reader had everything and could use almost none of it.

She'd come to learn how to actually run people analytics in her org — a real question, with a real next step waiting on the answer. What she found was abundance: hundreds of articles, a dozen frameworks, a stack of books, a search box that returned four hundred results ranked by nothing she cared about. Every piece was individually good. Collectively they were a wall. She did not need a four-hundredth thing to read; she needed the one thing to read next, given who she was and what she was trying to do — and that was the single thing the library could not tell her. She left with twelve open tabs and no first move, which is to say she left roughly where she came in.

This is the modern condition of any serious body of knowledge, and it has a precise diagnosis, written before the internet existed. The essay's claim is that the problem was never a shortage of content, and the cure is not more of it.

They say the answer is more, and better, content

The reflex of every publication, library, and knowledge base is to produce. More articles, more depth, more coverage, a better search box. The implicit theory is that the reader's problem is availability — that if the right thing exists and is findable, the job is done. So the work goes into making things and indexing them, and the measure of success is the size of the catalog and the relevance of the search.

But availability stopped being the bottleneck a long time ago. Herbert Simon named the real one in 1971, and it has only sharpened since: a wealth of information creates a poverty of attention.¹ When information is scarce, more is better and findability is the game. When information is overwhelming — which is now, for nearly every topic worth caring about — the scarce resource flips from the content to the reader's attention, and the job changes completely. The task is no longer "make the right thing exist and findable." It's "allocate this person's scarce attention to the highest-value next thing." A bigger catalog with a better search box is a better answer to the old problem, and no answer at all to the one we have.

The feed solved the wrong optimization

There is, of course, a machine that already allocates attention at scale: the feed. Recommendation engines decide what billions of people see next, so the problem of "what should this person attend to" is, technically, solved. Except the feed optimizes for the wrong target — it maximizes engagement, time-on-surface, the next click — and engagement is not advancement. A feed tuned to keep you scrolling is indifferent to whether you're any closer to running people analytics in your org; in fact it does better when you're not, because an unmet need is a returning user. The feed is a brilliant solution to attention allocation pointed at the platform's goal instead of the reader's.

So the open question isn't whether to recommend a next thing — it's what you're recommending toward. Toward more time on the surface, or toward the reader's actual progress? Those produce different software. One is a casino; the other is a guide.

What a guide needs that a feed doesn't

To recommend toward progress instead of engagement, a system needs three things the feed doesn't bother with, because progress is a harder target than attention.

It needs to know where the reader is — not their clicks, but their situation: who they are, what they're trying to accomplish, what they already know, what they've already tried, what just failed. That's a model of the person and their goal, not a log of their behavior.

It needs the content to be connected — not a pile of articles but a graph, where each piece knows what it teaches, what it presupposes, and what it sets up next. You cannot sequence toward a goal across a heap; you can only sequence across a structure. This is the unglamorous part — classifying what each thing is for and how the pieces relate — and it's the part the produce-more reflex always skips.

And it needs a direction — a theory of what "progress" even means here, so "next" has a gradient to climb. Education solved this generations ago: learning has an order, from knowing a thing to using it to judging with it, and you don't hand someone the advanced move before the foundational one.² A guide that knows the reader's level and the content's prerequisites can do what the library couldn't — hand her not the four-hundredth article but the first one that moves her, then the next.

Put those together and you get something closer to an adaptive guide than a publication: it reads where you are, matches the next-best piece from a connected body of knowledge, sequences toward an actual goal, and — the part that compounds — learns from what helped and what didn't, so the next reader's path is better than yours. The content barely changes. What changes is that the same library now spends your attention instead of flooding it.

The honest difficulty

This is harder than producing more, which is exactly why most knowledge surfaces don't do it. Modeling a reader's goal is harder than counting their clicks. Connecting a corpus — tagging what each piece teaches and requires — is slow, unglamorous work that produces no new content and so never feels like progress. And there's a real failure mode to avoid: a system that thinks it knows your goal and rails you down a path you didn't choose is worse than a search box, because at least the search box leaves you in control. The guide has to stay a guide — proposing the next step, showing its reasoning, and letting you overrule it — not a track you're locked onto. Adaptive has to mean responsive to you, not certain about you.

But the difficulty is the moat. Anyone can publish more. A body of knowledge that is genuinely connected — that knows what it teaches and in what order, and gets better at routing attention every time someone uses it — is a different kind of asset than a big archive with good SEO, and it can't be cheaply copied by producing more articles.

What changes when the library spends attention

Go back to the reader with twelve tabs. In the produce-more world she gets a better search box and four hundred good results, and the work of turning that into a first move stays hers — which means, for most people, it doesn't happen. In the adaptive world she states who she is and what she's after, and the library hands her the one next thing that moves her, then the next, adjusting as she reacts, and it remembers what worked so the reader behind her starts further along. The catalog didn't grow. The reader's attention finally got spent on her progress instead of drowned in the abundance.

Simon's line was a warning, not a lament: when information becomes wealth, attention becomes the thing worth designing for. We have spent two decades getting very good at producing information and at building feeds that harvest attention for the platform. The unclaimed move is to point that same machinery at the reader's progress — to build the library that spends your attention well. More content was never the answer. The right next thing always was.

This is a piece in the People Analyst program — the thesis under the adaptive-magazine engine and the connective substrate (classify → join → reveal): a body of knowledge that reads where you are, sequences the next-best step toward your goal, and learns. It's the consumer-facing face of the same connective layer that powers the communication engine; both are about matching, not producing. No figures are invented here.