Everything that has actually been read.

Every figure below is computed from the data files when this page opens. Nothing is hardcoded, which means nothing can drift upward while the underlying work stands still. If a number here looks small, it is small.

Counts

—

Loading the data files…These counts come from the same JSON the tools read, so they cannot disagree with what the tools show.

Comparison

The number a competitor would put here

350 sources. 500 million documents. 83 companies.

Those are real figures from real products in this category, and each of them is larger than anything on this page by two to seven orders of magnitude. There is no version of this where Day91 wins that comparison, and pretending otherwise would be the first unsourced claim on the site.

What those numbers count is coverage, which is a measure of how much was ingested. What this page counts is reading, which is a measure of how much a person opened, checked against the filing, and put their name to. The two are not the same quantity in different sizes; they are different quantities.

The useful question is not which number is bigger. It is which one you would be willing to repeat in front of a board. Take a claim from any of them, ask for the document, and see how many clicks it takes.

Bottleneck

Why it grows this slowly

A human approves every issue and every figure that enters coverage. That caps throughput at something like twenty to thirty companies, and the cap is not a stage the product will grow out of — it is the reason the output is worth anything.

In February 2026, Goldman Sachs and Yale published the Fin-RATE benchmark, which tested language models on the three things an analyst does: pull one fact from one filing, compare disclosures across companies, and track how one company's filings change over time. Accuracy fell by roughly 18.6% on the longitudinal task and 14.35% on the cross-company one. Those are the two tasks that produce any edge, and they are the two the models are worst at.

So the model here drafts and a person decides, coverage grows a company at a time, and the unread cells stay visibly unread. A product that let the model publish would have a much better-looking version of this page.

Growth

How a company gets added

Five readers ask for it The trigger is five separate requests for the same company. Not a vote, not a roadmap — five people who each independently needed it.

It files with the SEC No filings, no coverage. Private competitors can be tracked through The Mirror and The Diff, but they will never appear in this ledger.

Someone reads it end to end Adding a company is roughly a working day: the last two 10-Qs, the most recent 8-Ks, the pricing page, and the arithmetic checked twice.

It stays in only if it stays read A company whose figures go stale is worse than one that was never added, because the staleness is invisible. Coverage that cannot be maintained gets removed.

See every company in the index Read the method

Ask

Which company should be next?