Also: Federal Budget Transparency and AI-Powered Learning

AEIDataPoints-Banner ([link removed] )

Expert analysis made easy. Breaking down the news with data, charts, and maps.

Edited by Brady Africk and Lexi Baker

Happy Thursday! In today’s newsletter, we examine how victory in Iran would look, the executive branch’s annual budget, and how AI can improve learning.

Don’t forget—subscribe ([link removed] ) and send DataPoints to a friend!

For inquiries, please email [email protected] (mailto:[email protected]) .

1. Defining Success in Iran ([link removed] )

3.26_Pletka-Thiessen ([link removed] )

Topline: The US and Israel significantly degraded Iran’s air defenses and ballistic missiles in the first few weeks of Operation Epic Fury, reducing the country’s drone and missile strikes by 95 percent. Marc A. Thiessen and Danielle Pletka argue ([link removed] ) that even if the US stops short of toppling the regime, the campaign has already been a major success.

Three Phases: Following the destruction of Iran’s air defenses, the second phase of the war was the decapitation of the regime. Israel has taken the lead on this, successfully eliminating dozens of officials already. The third and final phase of the war—which has not yet been completed—is the elimination of Iran’s nuclear program.

The Timeline: President Trump suggested that US efforts could last for four to six weeks. If the regime has not fallen by then, Thiessen notes, Israel will likely continue the operations with US support.

“What we don’t see in the media is that the Iranians were upping the ante with every meeting. They weren’t just refusing to give up their nuclear program. They weren’t just refusing to give up their missile program. They were refusing to give up the fissile material, the hundreds of kilos of fissile material that they had.”

—Danielle Pletka ([link removed] )

More on the Iran War
([link removed] )

2. The Executive Branch’s Outdated Budget ([link removed] )

3.26_Capretta-1 ([link removed] )

Topline: In the 14 months since President Trump took office, the executive branch has released only one budget forecast. AEI’s James C. Capretta warns ([link removed] ) that releasing annual budget and economic data is a core responsibility of the executive branch.

The Lone Forecast: The administration’s only forecast—released in September 2025—was unusually short and raised many questions. Specifically, it did not use a standard baseline. Instead, it assumed a permanent extension of the 2017 tax cuts already passed in the One Big Beautiful Bill Act while excluding other spending cuts enacted in the same bill.

Budgetary Developments: Many developments have happened since September, and without a new budget, these could go unaddressed:

-

The Supreme Court ruled against President Trump in the tariff case, raising questions about future tariff revenue.

-

Congress rejected many of the administration’s domestic appropriations for 2026.

-

President Trump made it clear he wanted an increase in defense spending even before the war with Iran began.

“The federal government is not a static enterprise. A refreshed and complete budget and economic forecast from the administration at the start of each year is an important part of responsible and accountable government.”

—James C. Capretta ([link removed] )

More on Economic Data
([link removed] )

3. Can AI Improve Learning? ([link removed] )

3_26_Bailey ([link removed] )

Topline: AI is improving exponentially, advancing in two years from failing at basic math questions to correctly answering PhD-level questions. This leap raises concerns about students’ reliance on technology in school. AEI’s John Bailey cautions ([link removed] ) that AI can improve learning as long as it is designed to make students think harder rather than skip critical reasoning.

Design Is Key: AI prompts and systems must be structured to maximize positive student behaviors while minimizing negative ones. Research indicates that an unconstrained chatbot undermines learning, with one study showing that students using AI performed much better on practice problems but scored 17 percent worse once the AI was removed. However, with the right guardrails, the harm from AI-assisted learning largely disappears.

By the Numbers: In a recent study, students using an AI system that actively guided their learning—and did not just hand them final answers—achieved results equivalent to six to nine months of additional schooling, compared with peers who followed a fixed sequence of practice problems. These gains were driven by greater engagement—more time on task, more attempts, and higher-quality interactions with the AI.

“AI works in education when it’s used not to give answers but to orchestrate the conditions for productive struggle. The result is deeper engagement and stronger outcomes.”

—John Bailey ([link removed] )

More on AI in Education
([link removed] )

Dive into More Data

3.26_Carney-1 ([link removed] )

Can Washington Deliver Housing for Families? ([link removed] )

More on Housing
([link removed] )

3.26_McCusker ([link removed] )

The Cost of Operation Epic Fury ([link removed] )

More on Defense Spending
([link removed] )

Special thanks to Hannah Bowen, Drew Kirkpatrick, and Rosalie Blacklock.

Thanks for reading. We will be back with more data next Thursday!

For more data insights, subscribe today ([link removed] ) .

Email us (mailto:[email protected]) with data questions or ideas.

Ensure delivery ([link removed] ) | Subscribe ([link removed] )

American Enterprise Institute for Public Policy Research

Robert Doar, President

1789 Massachusetts Ave. NW, Washington, DC 20036

202.862.5800 | www.aei.org ([link removed] )

Unsubscribe from the link below. ([link removed] )

Facebook ([link removed] )

X ([link removed] )

LinkedIn ([link removed] )

YouTube ([link removed] )

Instagram ([link removed] )

Podcast Logo New_Gray ([link removed] )

Donate to AEI ([link removed] ) in support of defending and promoting freedom, opportunity, and enterprise.

This message is for: [email protected] | Manage preferences or Unsubscribe ([link removed] )