The short version
Four sources are read on every run. Two of them, Grants.gov and the Federal Register, are federal APIs that hand back structured records, and the rows that come from them are as accurate as the government’s own data. One, the State Justice Institute, is a website read directly because SJI publishes nowhere else. The fourth is a hand-assembled register of private, corporate and bar foundations, because private philanthropy has no catalogue and the two services that come closest both refuse automated access.
Every row carries a link to the page it was read from. That is the whole bargain: a row you can check beats a row that merely looks right. It matters more here than it did in AIJ’s leads tool, because a judge’s mailing address is wrong slowly and a grant deadline is wrong overnight.
Last built not yet.
What each source can and cannot see
Two kinds of row, and how to tell them apart
Read from a funder
A Grants.gov opportunity, a Federal Register notice, an SJI programme page. The title, the deadline, the award ceiling, the eligible applicant types and the description are the funder’s own, pulled from a structured field or from a heading on their own site. The subjects matched against them are the funder’s words and nothing else.
These are trustworthy to the extent the funder’s own record is. The usual failure is staleness: an agency leaves an expired notice up for months. Rows whose deadline has passed are re-labelled closed automatically, on every page load, however open the source still claims to be.
Researched into the register
A private, corporate or bar foundation. That this funder exists, gives to 501(c)(3)s, and has interests that reach AIJ is a judgement written into scripts/registry/foundations.mjs, not something scraped. Each entry carries a sentence saying what AIJ would actually ask them for, which is the most useful thing on the row and also the least verifiable.
What is checked on every run: that their page still answers, that it still reads like a grants page, any deadline or dollar range printed on it, and their most recent IRS Form 990. Because the subject match runs partly against AIJ’s own note rather than only the funder’s words, these rows have their mission score marked down a tenth, and every one says so.
How the score is built
One number out of 100, answering one question: if AIJ could go after exactly one more grant this quarter, how good a use of the time would this be? It is four components, and every row shows what each one scored and the sentence explaining why.
Mission fit
40How close the funder's own description sits to what AIJ teaches, or to somewhere AIJ said it would go. Read off the themes matched in the title and summary, weighted by how far out the ring is.
Can AIJ apply
22Whether a 501(c)(3) is named as an eligible applicant. The first question about any grant and the one most funders answer least clearly, so a full score needs the funder to have said it plainly.
Odds of winning
20How crowded the field is. Private and bar foundations start ahead of federal competitions, many awards beat one award, and an award far larger than AIJ's own programme counts against rather than for.
Ease of applying
18What it costs to get an application in: registrations that take weeks, a cost-share AIJ has to find, a full federal NOFO. Starts full and is spent down.
Grades
- A78 and upGo after this
- B63 and upWorth a serious look
- C46 and upOnly if the year is quiet
- D0 and upProbably not
What is deliberately left out of the score
The deadline
A closed grant that would have been perfect is worth more to AIJ than an open one that is a poor fit. Most federal programmes run again the following year, their guidelines are already written, and knowing about one in September beats finding it three weeks before next year’s deadline. Folding the date into the score would bury exactly the rows worth planning around, so timing is its own column and closed rows are kept in the data.
Size, as a good in itself
A bigger grant is not a better grant; a grant the size of AIJ’s actual programme is. Size enters through the odds of winning, where an award far above AIJ’s scale counts against rather than for, because reviewers of a multi-million-dollar award look for prior awards of similar size and that is a conversation a small nonprofit loses.
How far from AIJ a grant is allowed to be
The brief was explicit that a good grant beats a familiar one: AIJ would rather branch into a new subject with funding behind it than stay narrow without. So the subject taxonomy runs outward in four rings rather than drawing one boundary, and the outer rings are in the table rather than filtered out of it. The ring sets the mission score; it never removes a row.
What AIJ teaches now
0 rows- Judicial education and trainingTeaching sitting judges. This is AIJ's whole business and the closest a grant gets to it.
- Continuing legal educationCLE and CJE: the accredited hours judges and lawyers have to earn. AIJ's conferences are these hours.
- Court administration and operationsHow a court actually runs: caseflow, scheduling, court managers, the administrative office. AIJ teaches the judges who live inside these systems.
- Evidence, procedure and judicial reasoningEvidence, fact-finding, logic and decision-making from the bench. Four of AIJ's ten published knowledge areas are this.
- Judicial ethics and conductCodes of conduct, disqualification, judicial independence and the pressures on a sitting judge.
The bench and its courts
0 rows- Access to justiceSelf-represented litigants, plain-language forms, legal deserts. A judge's docket problem before it is anyone else's, and AIJ already teaches the self-represented litigant.
- Court technology and dataCase management systems, e-filing, remote hearings, court data. AIJ publishes 'Technology and the Courts' as a knowledge area.
- Judicial security and wellbeingThreats against judges, privacy of judges' personal information, judicial stress and fatigue. AIJ's 'Realities of Judging' covers the second half of this directly.
- Problem-solving and specialty courtsDrug courts, mental health courts, veterans courts, family treatment courts. Large federal money, and the judges who preside over them need training nobody gives them.
- Behavioral health and substance use in the courtsOpioids, serious mental illness and competency, as they land on a docket. Named an SJI Priority Investment Area for FY2026, which makes it the likeliest money in this ring.
- Family, domestic violence and elder justiceProtective orders, custody, guardianship and elder abuse: the dockets where a judge's training shows fastest. AIJ already teaches domestic violence under fact-finding.
- Tribal and rural courtsTribal justice systems and thinly staffed rural benches, both of which are under-served by every existing judicial education provider.
A pivot AIJ is open to
0 rows- AI, algorithms and the courtroomThe pivot AIJ named first. Generative AI in filings, algorithmic risk assessment, deepfake evidence, and what a judge is supposed to do about any of it. Funders here are new, well capitalised and short of credible partners who can reach judges.
- Juvenile justice and youth courtsThe other pivot AIJ named. Juvenile court judges, status offences, school-to-prison diversion, and the judges who decide what happens to a child. OJJDP is the largest single funder in this file.
- Underprivileged youth programsMentoring, after-school and out-of-school-time programs for children with the odds against them. Named in the brief as a direction AIJ would take. Judges are unusually credible messengers here, which is the angle a proposal would be built on.
- Civics education and the rule of lawPublic understanding of courts, civic education, judicial independence as a public good. A large and growing pot, and judges are exactly who these funders want in the room.
- Criminal justice research and reformSentencing, pretrial, reentry, bail, prosecution and defence. Judges sit at the centre of every one of these and are routinely left out of the training budget for them.
The right shape, a different subject
0 rows- Professional and adult educationCurriculum design, faculty development and continuing education for working professionals, whatever the profession. AIJ's competence is exactly this, and a funder of it is one conversation from funding a judicial version.
- Conferences and conveningsMoney for getting the right people into one room for three days, which is the literal thing AIJ produces. Conference support is a small line for most funders and an easy first grant.
- Nonprofit capacity and general operatingGeneral operating support, capacity building and organisational infrastructure. It pays for salaries and overhead, which the brief said is what a grant has to cover before AIJ will branch anywhere.
- Research and program evaluationStudying whether a program works. AIJ has twenty-five years of conference data and no published evaluation of it, which is a fundable gap and a credential worth having.
Why the funder type changes the score
The brief asked for private grants specifically, on the reasoning that there is usually less competition for them. That reasoning is right often enough to build on, so the funder type is not just a label: it sets the starting point for the odds of winning.
- Federal governmentA federal agency, applied for through Grants.gov. The largest awards in the table and the most competition for them. Needs a SAM.gov registration with a UEI, which takes weeks the first time.
- Federally chartered nonprofitCreated by federal statute, funded by Congress, run as a nonprofit rather than an agency. The State Justice Institute is the one that matters to AIJ: it exists to improve state courts, it funds curriculum directly, and its applicant pool is small.
- State government or judiciaryA state administrative office of the courts, judicial council, or state agency passing federal money through. Usually restricted to in-state applicants, which the eligibility column says.
- Bar foundation or IOLTAA state bar foundation or interest-on-lawyer-trust-account program. Small awards, an application that is a few pages rather than a few dozen, and a board made of lawyers who already know what AIJ is. The best effort-to-odds ratio in the table.
- Private foundationAn independent or family foundation giving from an endowment. Less competition than a federal competition and no SAM.gov registration, but many accept applications by invitation only, which is noted per row.
- Corporate or company foundationA company's own giving program or foundation. Fast decisions, smaller cheques, and an appetite for anything that reads as innovation. The AI funders in this table are mostly here.
- Nonprofit intermediaryA nonprofit that regrants somebody else's money, usually a federal agency's. Applying to one is a federal grant with the paperwork already done by somebody else.
The paperwork behind a deadline
A federal deadline three weeks out is usually not really three weeks. A first SAM.gov registration has taken anywhere from two to six weeks and no amount of good writing shortens it, and the Grants.gov account that depends on it cannot even be started until it is done. These are on every row that needs them, and they are what the ease-of-applying component is spent on.
- SAM.gov registration with a UEI2 to 6 weeks for a first registration, renewed annually
- Grants.gov account, with the AOR role authorised by the E-Business Point of ContactDays, and it cannot start until SAM.gov is done
- DOJ JustGrants accountDays. A second submission portal after Grants.gov, with its own later deadline
- eRA Commons account1 to 2 weeks
- SJI Grant Management System accountSame day
- NSF Research.gov organisation registration1 to 2 weeks
- The funder's own application portalUsually same day
- A letter of inquiry accepted before any full proposalNone. This is the low-cost way in
- By invitation only. Unsolicited applications are not acceptedNot applicable. The way in is a conversation, not a form
If AIJ has none of these yet, the sequence is SAM.gov first, then Grants.gov, then JustGrants if a Department of Justice programme is in view. Starting that today costs nothing and removes the reason most good federal rows in this table are not realistic this year.
What this tool will not do
Several funders and directories answer an automated request with a challenge page rather than with their content. That is an operator deciding they do not want to be read by a machine, and this tool treats it as the answer rather than as an obstacle. No challenge is solved, no headless browser is pointed at one, and no HTTP client is swapped to get a fingerprint a block list does not recognise.
- Candid, and the Philanthropy News Digest RFP feedThe closest thing to a catalogue of private grant opportunities that exists. Both answer with a Cloudflare interstitial. This is the single largest gap in the table and it is the reason the private side is a researched register rather than a sweep.
- americanbar.orgThe ABA Fund for Justice and Education, which would otherwise be a strong register entry. The related American Bar Endowment is readable and is in the table; the ABA route has to be checked by hand in a browser.
- The Annie E. Casey Foundation, and the Alfred P. Sloan FoundationBoth are in the register, because both are genuinely good fits, and both are flagged. Their rows are research rather than a reading of their sites, their pages are not requested at all, and everything on them should be confirmed in a browser.
One thing that looks like this and is not: a plain 403 from a firewall that a browser User-Agent fixes. That is a machine rejecting an unfamiliar string, not a publisher declining, and those hosts are named in scripts/lib/http.mjs and read with a browser User-Agent and a From: header carrying AIJ’s contact address. The same file also repairs one broken TLS certificate chain, sji.gov’s, by fetching the intermediate certificate from the address printed inside SJI’s own certificate. Certificate verification is never disabled.
Keeping it current
Deadlines move and programmes close. The table is a snapshot of the last run, and the overview page shows how old each source is with a coloured dot: green inside a week, amber inside a month, red past that.
npm run scrape- Read every source and rebuild the table.
npm run scrape -- sji foundations- Just those two. Takes under a minute.
npm run scrape -- --fresh- Ignore the six-hour disk cache and re-ask every server.
npm run verify- Ask every link in the table whether it still answers. This is the one that catches a funder who has quietly moved their grants page.
npm run audit- Read the rows rather than count them, and report values that cannot be right: a deadline in 1970, an award ceiling below its floor, a summary that is a cookie banner.
A source that cannot be reached keeps its last good rows and is marked failed rather than emptied. Grants.gov being down for an afternoon must never look like a week with nothing to apply for.