🛝 The Decision Came Before the Evidence

Welcome to Playground Post, a bi-weekly newsletter that keeps education innovators ahead of what's next.

This week's reality check: AI-like NIH proposals were associated with a four-point funding advantage while looking more like ideas that had already won. More than 1,900 colleges missed or underreported data meant to show what their programs deliver. And Indiana voters may decide the fate of nearly $200 million a year without seeing plans for most of the projected charter share.

💎 Data Gem

Only 49% of more than 357,000 New York City students said most or all of their teachers connect lessons to life outside school, according to the city's redesigned annual survey.

AI-Like Grant Proposals Won More Often

Federal grant reviewers are supposed to find the research most likely to move science forward.

The writing may be moving in a different direction.

A study published in the Proceedings of the National Academy of Sciences analyzed more than 125,000 funded, unfunded, and pending proposals submitted to the National Institutes of Health and National Science Foundation from 2021 through 2025.

Researchers used word-distribution modeling to estimate which proposals showed high levels of large-language-model involvement. They combined public award records with two confidential sets of full NIH and NSF submissions from two large research universities.

They did not observe or verify who actually used AI.

At NIH, proposals with high inferred LLM involvement were associated with a 4-percentage-point higher funding probability than proposals with low inferred involvement, according to the new grant analysis.

Successful NIH grants in the high-involvement group also produced 5% more papers.

But more papers did not mean more influential papers. The study found no advantage among the most-cited work.

Then came the deeper concern.

At both agencies, proposals with high inferred LLM involvement looked more similar to projects that had already received federal funding.

That could mean AI helped researchers write more clearly and align their proposals with reviewer expectations.

It could also mean the system became better at rewarding familiar ideas.

The study cannot say which explanation is correct, and it cannot show that AI caused the NIH funding advantage.

Both agencies already have policies on AI use.

NSF encourages disclosure and holds applicants responsible for accuracy and authenticity.

NIH has treated applications substantially developed by AI as a violation of its originality expectations since September 2025.

For education innovators, the opening is not another grant-writing assistant. Research offices need provenance systems that record where AI entered a proposal. Funders need similarity and novelty audits that show when a polished submission is converging on recent winners, plus review diagnostics that test whether writing style is quietly shaping portfolio decisions.

1,900 Colleges Missed the Data That Measures Their Value

More than 1,900 colleges either failed to report or underreported federal data on program cost, debt, and graduate outcomes.

They still have to finish the old requirement while preparing for its replacement.

The missing submissions cover the first two rounds of the Biden-era financial value transparency and gainful employment regulations. After several extensions, the Education Department gave institutions until January 15 to correct them or risk fines and other sanctions, according to the federal compliance notice.

A third and final reporting round is due October 1.

The transparency rule requires all colleges to report program costs and student loan burdens. The gainful employment rule adds outcome tests for career education programs, including a debt-to-earnings measure.

Then the regulatory yardstick changes.

A replacement earnings rule takes effect in mid-2027. Bachelor's programs will have to show that at least half of graduates earn more than a typical worker in the same state whose highest credential is a high school diploma. Graduate programs face a similar comparison with bachelor's degree holders.

Programs that fail the new test in two of three consecutive years lose access to federal student loans. A college can also lose Pell Grant eligibility in some cases if more than half of its federally aided students are enrolled in failing programs.

Those future consequences did not make the 1,900 colleges overdue. They belong to the replacement rule, not the older submissions.

But institutions must now operate both timelines at once.

They need to repair two past reporting rounds, prepare an October submission, decide whether to adopt parts of the replacement early, and build the data needed to model program risk before public results appear in 2027.

For education innovators, this is a program-level data infrastructure problem. Reporting pipelines can reconcile costs, loan burdens, completion, and earnings before submission. Regulatory crosswalks can keep the two rule sets separate, while portfolio dashboards flag programs approaching future federal-aid thresholds.

$48.7 Million Was Missing From the Plans Voters Will Judge

Eight Indiana school districts are asking voters to approve property tax increases worth nearly $200 million a year.

About 60 charter schools could receive an estimated $49.5 million of that total.

The law requires districts in four counties to share voter-approved revenue with charter schools that enroll students living inside their boundaries. In this year's election, the relevant referendums are limited to districts in Marion and Lake counties.

The money has not been approved.

Every dollar depends on voters saying yes in November. But the plans are already incomplete.

Only four of the eight districts included spending plans for individual charter schools in their state submissions, according to Chalkbeat's analysis of the potential revenue.

Of the projected charter share, $48.7 million was absent from the official spending plans.

That does not mean schools have no intended use for the money. Several charter leaders told Chalkbeat they would pay for staffing, transportation, professional development, attendance work, and student support.

It means the required district-to-charter handoff did not produce a public record before the vote.

Charter officials said some districts never asked for a plan. Some districts said charters did not provide one when asked. The process has little oversight, and responsibility is already being passed between organizations.

The operational stakes are real.

Christel House Academy South could receive $2.5 million if voters approve all three referendums in which it participates. The school transports more than 450 students, and its transportation costs have increased 28% in recent years.

Phalen Leadership Academies could receive nearly $6 million. Paramount Schools of Excellence could receive $4.1 million, and KIPP Indy could receive up to $3.6 million.

For education innovators, the gap is a shared referendum workflow that calculates allocations from student residency, assigns ownership for each required plan, and preserves an approval trail before the ballot. Public dashboards can then connect approved revenue to planned categories, actual spending, and the outcomes schools promised to improve.

⚡️More Quick Hits

This week in education:

• The FCC opened a rulemaking over an E-rate system that distributed $13.2 billion in five years — Nearly 25,000 entities received support from 2021 through 2025, and public comments on possible limits, conditions, or a sunset are due October 13.

• 79% of surveyed families eliminated at least one college because of price — A Sallie Mae-sponsored Ipsos poll of 2,000 students and parents put average family spending at $34,019, up 10% in one year.

• 57% of surveyed educators serving immigrant students reported student fear and anxiety — In the same subgroup, 39% reported increased absences and 18% reported enrollment declines tied to enforcement.

• CISA's workforce fell by nearly one-third as schools lost shared cyber support — MS-ISAC also lost federal funding and 70% of its membership before CISA issued two new K-12 security guides.

To stay up-to-date on all things education innovation, visit us at playgroundpost.com.