Counting the Zeros

Take one honest unit of work: the full-context read of a single company.

The unit

Ten years of filings. Every earnings call transcript. The patents, the hiring posts, the product reviews, the supplier chatter, the regulator's correspondence, the footprint in the trade press. Done properly, by a person, that read is three to five hundred hours. Fully loaded, call it fifty thousand dollars, plus six months of elapsed time that the market does not wait for.

That is the true price of knowing one company well. It is why desks concentrate. It is why watchlists are aspirational. It is why the tail of the market goes unread: not because anyone believes the tail is efficient, but because nobody could ever afford the unit.

Era	The same read	Cost	Elapsed
1985	Library, mail, microfiche	Thousands of dollars	Weeks
2000	Terminal and filings online	Hundreds	Days
2015	Cloud, search, alt data	Dozens	Hours
2026	Agent workforce	Dollars of inference	Minutes

The agent equivalent of that same read costs a few dollars of inference and one hour of senior review. Three orders of magnitude, give or take, inside a decade. We have watched this movie in compute, in storage, and in bandwidth. It is now playing in the one input investing never expected to get cheap: understanding.

The option book of ideas

When the marginal cost of curiosity falls to roughly zero, the question changes. It is no longer what can we afford to know. It is what is worth deciding.

Every investor carries a graveyard: the ideas that died because they were not worth a week of someone's time. Each of those ideas was an option. Cheap to explore, occasionally worth a fortune, and the desk could only ever afford to hold a handful at once. A human team holds dozens of live research threads. An agent workforce holds thousands, continuously, without fatigue and without resentment.

Alpha discovery stops being a bet on where to look and becomes a portfolio of perpetual looking. The math of that shift is brutal for whoever is on the other side of it: a desk holding a thousand cheap options will, over any honest sample of years, find things that a desk holding twenty cannot.

The tail

There are roughly four thousand listed names in the United States alone, and coverage is grotesquely concentrated. The median large cap is read by everyone. The median small cap is read by almost no one. Inefficiency lives where attention isn't, and attention was simply too expensive to send down the tail.

The most mispriced securities in the world are not hidden. They are unread.

Scarcity pricing works in reverse. When the cost of knowing falls by three orders of magnitude, the value of having asked the question goes up. The desks that internalize this first will spend the next decade reading the parts of the market the rest of the industry still cannot afford to visit.

The next essay is about what the desk itself becomes when capacity turns elastic.