About
FilingSmith is a product built and operated in Australia. It reads US securities filings as they are published, computes what changed against the prior comparable period, and delivers the difference — with every figure traceable to SEC’s own data — to the people who follow those companies.
Why it exists
Public companies disclose an enormous amount, on a schedule, in a standard format, for free. The information is not hidden. It is just long, and it arrives while you are doing something else.
The gap this product addresses is narrow and specific: between a filing was published and you know what changed in it, for the ten to forty companies an individual investor actually follows. Not research, not recommendations, not a data terminal. A diff, pushed, with its sources attached.
How it is built
The design decision that shapes everything else is that numbers are computed, not generated.
Financial figures come from SEC’s XBRL APIs — the already-structured, as-reported data SEC publishes for every tagged value in a modern filing. Period-over-period deltas are arithmetic on those values. A language model is used only to write the sentence around figures the engine has already fixed, and its output is checked afterwards: any figure that cannot be reconciled against the computed diff causes the whole narrative to be discarded rather than repaired.
It is also the only way we were willing to ship a product whose entire value is that you can believe the numbers.
That is a slower and more expensive way to build this than handing filings to a model and asking what changed.
How it is run
FilingSmith is independently owned and funded. It takes no payment, commission, or consideration from any issuer, broker, exchange, or data vendor for what it reports, and it has no relationship with the companies whose filings it processes. Filings are public; the product is paid for by the people who read them.
It is a small operation — deliberately narrow in scope, with the engineering effort concentrated on the diff and the checks rather than on breadth.