Federal Funding Routes

What this is, and what it is not

A registry of 23 US federal funding routes — 17 non-dilutive, 4 debt, 1 equity, 1 hybrid — each recorded from its own programme page with the URL it was read from and the date. The screen above compares what you enter against the eligibility criteria recorded on each record and sorts the registry into eligible, not eligible and undetermined.

It is not a determination that any route will fund a company. Agencies decide that at review, on a full application, against criteria a registry cannot hold — reviewer discretion, competing applications, the fit of the work to the call. The tool does not rank routes, does not score them, and does not pick one. Which capital to take changes control, runway and personal liability, and that is a decision with a person's name on it.

Why "undetermined" outranks "eligible"

One unanswered question makes the whole route undetermined, however many other criteria pass. A company that has not said whether it is incorporated has not said that it is. A screen that treats silence as a pass sends someone to spend three weeks on an application they were never going to win, and that is the failure this design exists to prevent. Not sure is the default on every question for the same reason, and a blank number is unstated, not zero.

A hard disqualifier is a real answer and saves time. A silence is not an answer, and is never counted as one.

Where the figures come from, and what "read by a machine" means

Each record cites one programme page. The figures on it — a ceiling, a rule, a deadline — were transcribed from that page by an automated process on 2026-09-01, then re-checked by a second automated pass that fetched the same page and confirmed each figure was actually on it. Figures that were correct but lived on a different page were removed rather than mis-cited. Where a page stated nothing about timing, the deadline is recorded as not stated rather than as rolling.

That is better than a figure from memory and worse than a person having read the page. A machine can transcribe the wrong row of a table, or miss a footnote that makes a number conditional. As of this snapshot 0 of 23 records have been read by a person. The page says so in its chrome, above the results, in the disclosure and on every route, and the counts come from the data rather than from this sentence — so when that changes, the page changes with it.

Ceilings are stated per source page

Several NIH funding announcements list a different ceiling for each participating institute. What is recorded is the highest ceiling on the page, and a company applying to a different institute faces a different number. So the card says ceiling on the source page rather than maximum, and the disclosure says why. Read the page for the institute you are applying to.

Freshness runs on your clock

Grant windows close and rate sheets move after a site is built. Each route carries the date its source was read and a recheck interval; the screen evaluates both against your own clock, so a route whose deadline passed yesterday is withheld today — shown by name with the reason, never as an open route — without waiting for anyone to rebuild the page.

Why we are giving it away

Because it is public information behind a dozen agency pages in a dozen formats, and assembling it with provenance is an afternoon nobody should spend twice. The discipline here — every figure carries its source and its date, and the tool refuses to make a decision that belongs to someone else — is the same discipline in the DHF gap reports we charge for.

Questions we get

Does this tell me whether we will get funded?

No, and it is built not to. Each result is a comparison of what you entered against criteria recorded from the programme page. Whether a company is funded is the agency’s decision at review, and the tool does not rank routes or pick one. An unanswered question is reported as unknown, never as a pass.

What does "read by a machine, not yet confirmed by a person" mean?

Each figure was transcribed from its source page by an automated process, then re-checked by a second automated pass against the same page. As of this snapshot 23 of 23 records are in that state and 0 have been read by a person. A machine reading the page is better than a figure from memory and worse than a person reading it: it can transcribe the wrong row of a table or miss a footnote that makes a number conditional. Every route links to its page so you can read it yourself.

Why does the ceiling say "on the source page" rather than "maximum"?

Because for some programmes it is not a programme-wide maximum. Several NIH announcements list a different ceiling for each institute; what is recorded is the highest on the page, and a company applying to a different institute faces a different number. The label says where the figure came from, not what it guarantees.

Why are so many deadlines "not stated"?

Because the programme page states nothing about timing, and recording "rolling" would be an inference rather than a reading. 13 of 23 routes are in that state. Agency loan programmes in particular are delivered through lenders on the lender’s timetable, which no SBA page states.

What happens to a route when its deadline passes?

It moves to "Withheld today" on the day the deadline passes, evaluated against your own clock rather than the build date. It is shown by name with the reason, and never as an open route. The same happens when a record’s source verification is older than its recheck interval.

Do you store what I enter?

No. Nothing you enter is stored, sent, or logged. The analytics events carry the tool name and which group a route was in when you opened it, and the transport drops anything else. A company’s ownership, revenue and headcount are its own business.