The Wrong Scarcity
Herb Simon, 1971: “A wealth of information creates a poverty of attention.”
That sentence is fifty-five years old and we’re still building systems that ignore it.
Here’s the pattern. Some resource is scarce for a long time. People build tools to produce more of it, move it faster, waste less of it. The tools work. The resource becomes abundant. And then something strange happens: the scarcity doesn’t disappear. It migrates. It shows up somewhere else in the system — usually in whatever was previously unconstrained because nobody needed to think about it.
The tools don’t migrate with it. They keep solving for the old scarcity. Sometimes for decades.
Information used to be scarce. Libraries, journals, expert networks — the whole knowledge economy was organized around the problem of finding things out. Then search engines, digitization, open access, and finally generative AI made information essentially free. The scarcity migrated from information to the attention needed to process it. But we still build tools that maximize information delivery. Your inbox, your feed, your notification tray — all engineered to move more information to you faster. They’re solving a scarcity problem from the 1980s while creating an attention crisis in the 2020s.
Code used to be scarce. Writing software was expensive — skilled labor, long timelines, careful specification. The open-source movement and now AI coding assistants have made code production nearly free. The scarcity migrated from writing code to maintaining it. But the ecosystem still celebrates contributions over stewardship. GitHub counts commits, not code reviews. Hackathons reward building, not pruning. We measure developer productivity in lines produced, not lines prevented. The old scarcity was supply. The new scarcity is attention to what already exists.
Model capability used to be scarce. Two years ago, the question was whether AI could write coherent paragraphs, pass exams, generate working code. Now dozens of models can do all of those things. The scarcity migrated from capability to judgment — knowing when to use which model, how to frame the prompt, when to trust the output, when to verify. But the industry still benchmarks on capabilities. Leaderboards measure what models can do. Nobody benchmarks how well humans can direct them. The expensive part moved from the model to the operator, and we’re still optimizing the model.
Bandwidth used to be scarce. The entire telecommunications infrastructure was designed to move bits from point A to point B. Now bandwidth is effectively unlimited for most purposes. The scarcity migrated to relevance — not “can this message reach you?” but “should it?” Every messaging platform, every collaboration tool, every social network was designed to reduce the cost of sending a message. None were designed to reduce the cost of deciding whether to read one. Slack gives you seventeen channels and a notification for each. It solved the 1990s problem of “how do we stay connected?” and created the 2020s problem of “how do we stay focused?”
The lag between scarcity migration and tool migration is the gap where most operational pain lives.
When an engineer complains about “alert fatigue,” they’re experiencing a system that was built to produce alerts (scarce information about system health) but never redesigned for the world where alerts are abundant and attention is the bottleneck. The fix isn’t better alerts. It’s better filtering.
When a team complains about “meeting overload,” they’re experiencing a tool (the calendar invite) designed to solve coordination scarcity — getting the right people in the same room. In a world where anyone can schedule anyone, the scarcity is no longer coordination. It’s uninterrupted time. The fix isn’t shorter meetings. It’s a system that protects attention rather than consuming it.
When an AI researcher complains that their multi-agent system produces impressive individual outputs but can’t synthesize them, they’re experiencing coordination tools designed for the physical world — where moving information between agents is hard — applied to the digital world, where moving information is free but integrating it costs context. The fix isn’t better pipes between agents. It’s better filters on each agent’s intake.
The wrong scarcity is a trap because it feels productive to solve. The tools are familiar. The metrics are established. The infrastructure is built. Solving for the old scarcity generates visible output — more messages sent, more code written, more alerts fired, more information surfaced. The new scarcity is harder to measure. “How much attention was preserved?” doesn’t have a dashboard. “How many irrelevant inputs were filtered before they consumed context?” isn’t a KPI anyone tracks.
The pattern isn’t just about information. Every time a fundamental constraint shifts, the same thing happens. When food production stopped being the binding constraint on population, we didn’t suddenly have no problems — the constraint migrated to distribution, then to nutrition quality, then to metabolic disease from overabundance. When physical safety stopped being the primary concern of daily life, anxiety didn’t disappear — it migrated to social, financial, and existential threats. The scarcity always moves. The question is how long it takes us to notice.
How to spot it: look at any system that feels effortful despite having “enough” of something. Enough information but can’t find what matters. Enough tools but can’t choose. Enough communication but can’t coordinate. Enough output but can’t integrate. The “enough” is the signal. It means the old scarcity was solved. The effort you’re still feeling is the new one.
Then look at what the system optimizes. If it’s still optimizing for supply of the thing you have “enough” of, you’ve found the lag. The fix is almost always the same shape: stop building pipes and start building filters.
The hard part isn’t the engineering. It’s admitting that the expensive infrastructure you built to solve the last scarcity is now contributing to the next one.