IFP Roundup: DOGE, AI action plans, supersonic airliners, and much more

It’s been a busy month! Caleb wrote a piece for The New York Times on what “government efficiency” should mean for science. Santi went on The Ezra Klein Show to talk about DOGE. We submitted a comprehensive AI Action Plan to OSTP. Tim and Arnab also hit the podcast circuit to discuss Compute in America — first at Lawfare, then on ChinaTalk. And Brian taught everyone about solar energy, jet engines, and supersonic flight.

One quick announcement: we’re hiring a full-time or part-time Operations Manager to handle office and event management. We will pay a bounty of $3,000 for a referral that leads to a successful hire.

1. What “government efficiency” should mean for science

In an op-ed for The New York Times, Co-founder Caleb Watney argued for a better approach to science funding amid DOGE's cost-cutting:

• Our scientific system misses out on valuable research: Despite eventually winning a Nobel Prize and enabling COVID-19 vaccines, Katalin Karikó's pioneering mRNA research was repeatedly denied grants and delayed by a decade due to risk-averse funding systems.

• DOGE's current approach could make things worse: While our system needs reforms, the administration's 15% cap on indirect costs for NIH grants could lead to $4 billion in annual biomedical funding cuts, forcing universities to lay off researchers and close labs.

• "Efficiency" should be about ROI: Instead of equating efficiency to cost-cutting, we should maximize returns on scientific investment. We can do this by paring back regulations and supporting high-risk, high-reward research.

• We need new ways to fund science: Cutting-edge science is increasingly performed by organizations beyond traditional academia. The NIH should develop new funding mechanisms to support large-scale scientific infrastructure.

Read the full piece here.

2. What is DOGE's real goal?

Senior Editor Santi Ruiz went on The Ezra Klein Show to discuss his “50 Thoughts on DOGE” essay. Here’s a taste of some of those thoughts:

• Twitter is an adversarial information environment: Elon Musk is an unreliable narrator, and the feedback loops he’s constructed on Twitter have shaped DOGE in unexpected ways.

• DOGE's focus has shifted: Although it was initially pitched as an agency for targeting overregulation, DOGE now emphasizes headcount reduction and dollar savings at the risk of reducing government effectiveness.

• Staff cuts target the wrong people: Due to civil service rules, DOGE is primarily firing probationary employees — often younger, techier workers who could improve the way government functions.

• Multiple agendas are at work: Some DOGE initiatives would have happened without it, as they align with longstanding conservative goals to reduce specific agencies or reassert presidential control over spending.

Read all 50 thoughts here. Watch Ezra Klein’s interview with Santi here.

3. An action plan for American leadership in AI

IFP submitted recommendations to the Office of Science and Technology Policy for its AI Action Plan RFI. US policymakers should:

• Establish Special Compute Zones: Designate regions where AI clusters of at least 5 GW can be built rapidly through coordinated federal and private action, while streamlining permitting and enhancing security requirements.

• Support American open-source AI: Launch prize competitions to incentivize the development of reliable open-source AI models, and provide free hosting of American open-source models for small developers via the National AI Research Resource.

• Fund ambitious research moonshots: Invest in a "grand challenge" for AI interpretability, tamper-resistant hardware security, and AI-driven formally verified software to overcome key reliability and security challenges.

• Build state capacity for model evaluation: Reform the AI Safety Institute to focus on national security risks and directly report to key national security decision-makers about the capabilities of both US and adversary AI models.

• Improve export controls and enforcement capacity: Use export controls as market-shaping incentives to develop more secure AI chips, restrict access to high-performance inference chips, and modernize BIS.

Read the rest of the plan here. Kevin Roose and Casey Newton also discussed our submission on their Hard Fork podcast — listen here.

4. How to strengthen America's AI workforce

In another submission to OSTP’s AI Action Plan RFI, our Director of Immigration Policy Jeremy Neufeld worked with Matthias Oschinski, Jacob Feldgoise, and Sophie Alcorn to pull together 11 recommendations to recruit the world's top AI talent. The US government should:

• Launch Project Paperclip 2.0: Initiate a targeted talent recruitment program to attract top AI talent for national defense purposes, similar to past recruitment programs.

• Add premium processing for entrepreneurs: Add premium processing to the International Entrepreneur Rule program to support AI startup founders and reduce processing times.

• Cut red tape: Update the Schedule A shortage occupation list to include AI fields, implement special handling for advanced STEM talent, allow domestic visa renewal, and deploy AI for visa processing.

• Support AI talent: Clarify dual intent for O-1A visas, recapture unused green cards, authorize work for O-3 spouses, launch a targeted talent recruitment program, and provide export control clarity for immigrant inventors.

Read the full report here. Jeremy also spoke about the role of high-skilled immigrants across the defense sector on the Emerging Tech Horizons podcast.

5. Understanding solar energy from first principles

Senior Infrastructure Fellow Brian Potter explained the dramatic rise of solar power and evaluated its long-term potential:

• Solar is growing at unprecedented speed: Solar capacity has expanded faster than any energy source in history, growing from 100 to 1,000 terawatt-hours in just 8 years. It now represents about 50% of planned new electrical generation capacity in the US.

• Costs have fallen dramatically: Since its invention in the 1950s, solar PV costs have decreased by a factor of nearly 10,000. In the last decade alone, cell costs dropped by more than 50%, making solar one of the cheapest electricity generation methods.

• Intermittency remains the key challenge: Unlike fuel-based power sources, solar can't generate electricity on demand. Day-night cycles, seasonal changes, and weather conditions significantly impact solar output.

• Overbuilding can address intermittency: Meeting high percentages of electricity demand requires "overbuilding" — installing excess panel capacity and storage to compensate for periods without sunlight, which becomes economically viable if costs continue to fall.

Read the full analysis here.

6. How to build compute in America

Director of Infrastructure Policy Arnab Datta and Director of Emerging Technology Policy Tim Fist discussed how to build out AI infrastructure in America on the Lawfare and ChinaTalk podcasts:

• Multiple energy sources will be needed: We recommend an "all-of-the-above" approach including natural gas in the short term, solar with battery storage where feasible, and next-generation geothermal and small modular nuclear reactors as they become available.

• Government must reduce uncertainty: The greatest obstacle to financing next-generation energy projects isn't quantifiable risk, but true Knightian uncertainty — unpredictable technical, supply chain, and regulatory variables that defy probabilistic modeling and pricing by conventional financiers.

• Security requirements should accompany federal assistance: As AI companies build powerful systems that could reshape global power dynamics, the government should require robust security measures at data centers to protect against theft of model weights and sabotage.

• Geothermal emerges as a breakthrough energy source: Next-generation geothermal energy could unlock vast resources by applying fracking techniques to create artificial reservoirs in the Earth's crust, providing abundant, reliable, and clean power for AI data centers.

Listen to ChinaTalk here and Lawfare here.

7. A solution to the housing shortage

Non-Resident Senior Fellow Christopher Elmendorf and Alex Armlovich wrote a piece for City Journal on their proposal to reform the low-income housing tax credit to boost housing supply:

• Focus on cities, not suburbs: Rather than attempting politically difficult suburban zoning reform, Congress should target big, expensive cities where voters already support more housing development but local interest groups block progress.

• Reallocate housing tax credits: The $15 billion federal affordable housing tax credit program should be withheld from major cities unless they adopt key pro-housing reforms, redirecting those funds to other areas within the same state.

• Required reforms for cities: To remain eligible for tax credits, cities would need to allow apartment buildings along major corridors and in commercial areas, cap development fees at $10,000 per unit, waive price controls that make projects infeasible, and approve projects “ministerially.”

Read the full op-ed here.

8. How to run a private military company

Santi interviewed John Lechner about his new book Death Is Our Business: Russian Mercenaries and the New Era of Private Warfare on the Wagner Group:

• Wagner evolved over time: Beginning with Russian involvement in Ukraine in 2014, Wagner Group expanded from military operations to diplomacy, mining, brewing, and other enterprises in regions like Syria, Libya, and Central African Republic.

• PMC models vary significantly: Unlike Western contractors like Blackwater that primarily secured facilities or personnel, Wagner often engaged in direct offensive military operations and expanded into diplomatic and economic activities.

• Wagner wasn’t just an arm of the Russian state: Wagner functioned both as a tool of the Russian government and as entrepreneur Yevgeny Prigozhin's personal business venture, with varying degrees of state oversight in different regions.

Read or listen to the full interview here.

9. Why it’s so hard to build a jet engine

Brian explained why commercial jet engines represent one of civilization's toughest technical challenges:

• Performance and economy are tightly linked: Jet engines must be fuel-efficient, require minimal maintenance, operate at extreme temperatures, and remain lightweight — all while lasting thousands of flight hours.

• Only a few organizations can succeed: Just three companies (GE, Pratt & Whitney, and Rolls-Royce) manufacture large commercial jet engines, and developing a new engine costs billions of dollars.

• Technical boundaries must be pushed: Each new engine project requires companies to further optimize compression ratios, materials, and operating temperatures in key technologies. The RB211 engine drove Rolls-Royce into bankruptcy before becoming successful.

Read the full piece here.

10. Will Boom successfully build a supersonic airliner?

Brian also examined Boom Supersonic's recent milestone of breaking the sound barrier with its XB-1 test aircraft and evaluated the company's chances of successfully building a commercial supersonic airliner:

• Technical hurdles are enormous: Building commercial aircraft and jet engines are already among the most difficult engineering challenges. Boom is attempting both simultaneously, with the added complexity of supersonic flight.

• XB-1 success is just the beginning: While impressive, Boom's supersonic test flight doesn't address the harder challenge of building an economical, reliable airliner that meets performance targets.

• Engine development is a major risk: After established manufacturers declined partnerships, Boom is designing its engine in-house — a highly unusual and risky approach for an aircraft manufacturer.

• The business case remains uncertain: Even if technical challenges are overcome, the market for supersonic travel must be large enough to justify development costs of approximately $10 billion or more.

Read the complete assessment here.

11. Links from friends of IFP

• Max Tabarrok offers one clear metascience reform: Rely less on external academic reviews for grants and more on (faster) internal reviews.

• Dwarkesh Patel wrote a book for Stripe Press about the scaling era in AI.

• Matt Clancy takes a step back and reviews his own life as a way to think about the value of technological progress.

• Open Philanthropy is launching the Abundance and Growth Fund, and is seeking a Senior Program Officer to lead its strategy and grantmaking, with at least $120 million committed over the next three years. They’re looking for an experienced leader to fund high-impact opportunities and cultivate a thriving ecosystem of researchers, policymakers, entrepreneurs, and funders. Learn more and apply here by March 31st. Refer a strong candidate here and get a $5k reward if Open Phil hires them.

👋 Tweet for the road