Editorial Policy

Essays published in /thinking are drafted by an in-house pipeline that uses large language models. They are bylined to Go Fund It Now, the organization, because the organization is accountable for them. They are not presented as the personal writing of any individual, and no essay carries a human byline it did not earn.

We state this because it is true and because you would be entitled to be annoyed to discover it later.

Each piece moves through four passes before it can be published:

1. Angle. A thesis is selected against a fixed set of arguments the company actually holds. The recent published essays for that argument are shown to the model with an instruction not to repeat them. That is guidance, not verification: it makes a duplicate less likely rather than impossible.
2. Draft. The essay is written under constraints that forbid inventing statistics, dollar figures, named organizations, quotes, or dates.
3. Critique. A separate pass reviews the draft and flags suspected fabrications. A flagged draft is regenerated. A draft that still carries flags is never published automatically.
4. Packaging. Title, description, definition and questions are produced and validated for structure.

A piece publishes automatically only if it clears the critique threshold, carries no fabrication flags, and does not collide with an existing essay's URL, which is how we detect that the same argument has been made before. Anything else is held as a draft for a human to review, reject, or edit. A re-argued thesis that happens to produce a different URL is not caught by that check.

It has missed things, and saying so is the point of this section.

The safeguards above operate on one piece at a time. They are good at catching an invented number and poor at catching a true-sounding claim that nobody measured, or a regulatory figure that was correct once and has since changed. In August 2026 an essay of ours stated a federal Single Audit threshold that had been superseded nearly two years earlier. It was corrected when found. The generator has also repeated an unsourced seasonal claim across several essays before we noticed the pattern.

We publish this list rather than a list of safeguards alone because the failures are the honest measure of the process.

Where a piece states a figure, it should carry its source and the date it was accurate. We prefer primary sources: the Code of Federal Regulations, grants.gov, the issuing agency, the funder's own notice. Statistics the platform surfaces about our own feed are described in our methodology.

Nothing in the library is legal, financial, or compliance advice. Regulatory thresholds change. Verify against the primary source before acting on anything you read here.

We correct in place rather than quietly reissuing. A corrected essay carries a visible last-updated date and a machine-readable modification date, and the correction keeps the original figure in view where the change is material, so a reader can see what moved.

To report an error, write to [email protected] with the URL and the claim. Factual corrections are made regardless of whether the error favours us.

Essays are not sponsored and no funder, vendor, or client pays for coverage or placement. The library exists to argue a position we hold about how funding work should be done, which means it is opinionated by design and should be read that way.