The Last Scarcity

Every era of investing is defined by whatever is scarce.

In 1975 the scarce asset was information. If you wanted to know what a company had earned, you waited for the annual report to arrive by mail, or you got on a train and went to look at the factory. The edge belonged to whoever did the legwork, and the legwork was brutal. Graham called it security analysis. Everyone else called it work.

In 1982 Bloomberg put a terminal on the desk and the scarcity moved. Data became cheap and speed became dear. Through the nineties and the two-thousands the quants industrialized the trade: better models, faster pipes, tighter execution. By the twenty-tens the scarcity had moved again, this time to exhaust. Satellite parking lots, card panels, app downloads. Everyone with a budget could buy the same feeds, so the edge went to whoever asked the data better questions.

Notice the pattern. Alpha never disappears. It migrates to the new scarcity, pays the people who arrive first, and hardens into infrastructure behind them. Then it moves again.

Today data is effectively infinite. Compute is effectively free. Models read, write, and reason at a level that would have made a research director in 2015 quietly close his office door. And yet the desk feels more overwhelmed, not less. Earnings season is still triage. The watchlist is still a wish list. Somewhere in the backlog sits the idea that would have made the year, unread.

The scarcity moved again. Almost nobody has repriced it.

The scarce asset now is attention. Not data, not speed, not even intelligence in the machine-learning sense. Attention: the serial, sleeping, forgetting, gloriously human capacity to point a mind at something long enough to understand it. Every investor I know is rich in information and poor in the one thing that turns information into conviction.

This series is about what happens when that scarcity ends. Not because humans get smarter, and not because machines replace them. Because for the first time in the history of the industry, attention itself can be manufactured.

The claim, stated plainly: AI agents are the first technology that scales attention rather than computation. Every tool before them, from the ticker to the terminal to the cloud, made an investor faster at executing decisions already taken. Agents operate one layer higher, on the reading, watching, noticing, and remembering that decisions depend on. They do not decide. They make sure that nothing worth deciding goes unseen.

What that reprices is the subject of this series. Essay I takes apart the investment hour and shows where it actually goes. Essay II explains why attention never scaled, and what an agent actually is. Essay III counts the zeros in the collapse of research costs. Essay IV describes the elastic desk: cognitive labor as a utility. Essay V asks the only question that matters, which is what survives when everyone can know everything. Essay VI takes the objections seriously. Essay VII is the window: what to do, and by when.

Each essay stands on its own, though the series is best read in order.

A note on incentives, because incentives matter. I work on this problem. Konclave builds agent workforces for institutional investors, and you should discount my enthusiasm accordingly. The argument does not depend on Konclave. It depends on arithmetic, and arithmetic does not care who builds it.