Data centers vs. the grid, Houses atop post offices, and Dispatches from Marblehead
What we’re reading, May 28, 2026
We’re glad to be back after a short break! Here’s what caught our attention over the last couple weeks:
The Jones Act, the American law that bans ships manufactured outside the US from traveling between US ports, has been an economic burden on the country for decades. The best study on its impact estimates the cost at 0.1 to 0.4 percent of GDP — which may sound small, but given the size of the US economy that means $31.8 billion to $127.4 billion in economic damage every single year. It’s particularly bad for clean energy, because offshore wind turbines need specialized ships to install them, and the US has built a grand total of one of those ships. Luckily, the Trump administration has waived the Jones Act for oil, fertilizer, and related products since March as part of the ongoing Iranian oil crisis. Colin Grabow at the Cato Institute pulled the data and found that in just two months, 45 voyages from 35 ships have taken place under the waiver, mostly moving various oil products from the Gulf Coast of Texas and Louisiana to the West Coast and Puerto Rico. That suggests meaningful pent-up demand for transit using foreign-built ships that the Jones Act is suppressing. — Dylan Matthews
Earlier this month NSF announced its new X-Labs program, a $1.5 billion, decade-long initiative to fund independent, milestone-driven research teams working outside of traditional university and corporate structures. The program aims to address a long-held concern that existing grant mechanisms and institutional models aren’t well suited for team-oriented, engineering heavy, long-term research efforts. This is a relatively small bet, all things considered (it’s ~1% of NSF’s annual budget); but it’s an exciting one that will test a new model for doing science, and support a new ecosystem of research labs taking big swings. — Jordan Dworkin
In other news, arXiv’s CS section chair Thomas Dietterich announced that authors who submit papers with “incontrovertible evidence” of unchecked AI content (e.g. hallucinated references or leftover chat content like “here is that manuscript you requested”) will be banned for a year. After that, future submissions will be required to clear peer review before being posted on arXiv. This caused a bit of a stir online, with some researchers supportive and many strongly opposed. One opposition camp was frustrated that this move shifts arXiv even further away from the openness and lack of gatekeeping that made it valuable in the first place and is needed going forward. I’m sympathetic to that concern, especially with regard to the post-ban requirement that submissions pass peer review. But another camp seems to find it unreasonable to expect authors to verify their claims and citations, which I am less sympathetic to. It is certainly true that references have included mistakes or misattributions since long before AI, and these are usually relatively harmless (though they don’t suggest great citation practices on the part of authors). But it’s never been good when the knowledge ecosystem is flooded with unsourced claims, and AI allows that to happen at speed and scale previously unthinkable; if researchers can work with AI to write a paper that is accurate and rigorous, fantastic! But having AI write an inaccurate paper that is never carefully reviewed is not fantastic. Holding authors accountable for the accuracy of the work they’ve put their name on seems worthwhile. — Jordan Dworkin
It seems obvious that increased demand for electricity from data centers must raise prices. But electricity prices can move in counter-intuitive ways. We’ve talked before about how electricity price increases largely hadn’t yet been empirically linked to new demand (read: data centers) in most geographies outside the PJM Interconnect (the electricity market that houses data center alley). And in places like North Dakota, increased demand for electricity has been associated with lower electricity prices, since fixed costs of infrastructure could be spread over a larger ratepayer base. Travis Kavulla (incidentally, potentially the next head of the Bonneville Power Administration) explains in American Affairs this week why more places will probably look like PJM - where the grid has little excess juice to squeeze, and new demand mostly does require costly new investment - than like North Dakota as data center demand continues to rise. Additionally, while grid capacity can at least in theory cut both ways, equipment price inflation and higher capital costs really only drive up the cost to serve demand, relative to existing customers. Kavulla explains why bringing on new demand at the same rates as existing customers - whether through typical ratemaking or longer-term take-or-pay agreements - means socializing those higher costs across all utility customers, and argues that treating large loads like other customers doesn’t make sense when they’re potentially driving multiples of existing demand. To avoid the malignancies of the famously strangled generation queue on the load interconnection side, he argues, among other propositions, for an “open season” for large load interconnection to manage grid access, together with strict BYOG (Bring Your Own Generation) requirements that actually match the cost and risk of new supply to new demand. — Willow Latham-Proenca
Pew’s Housing Policy Initiative is making a splash with its new video explaining “address chain” filtering: How new market rate construction frees up housing opportunities across incomes and neighborhoods as occupants of newly built homes free up their existing, cheaper homes (not over decades but in just months). Pew also launched a new report on pre-approved building plans — reusable, pre-cleared design blueprints that let builders bypass discretionary design review. The report documents results from ~40 US jurisdictions running some version of the program, with the strongest outcomes in small and mid-size cities like South Bend, Indiana (223 new homes built, ~7% of permits over three years) and Seattle (ADU approval times cut from 160 days to 54). Federal interest is growing: the Accelerating Home Building Act, folded into the 21st Century ROAD to Housing Act, would launch a HUD pilot grant program to scale the model. — Alex Armlovich
“The USPS owns over 8,500 properties and leases over 25,000 more. It also needs money.” In a new Brooking Report, Aaron Shroyer, Ben McAdams, and Glen Nuckolls argue we can kill two birds with one stone: the Postal Service should let developers build housing above post offices (or otherwise use their real estate to support housing). They estimate that rebuilding Postal Service real estate with buildings that merely have the same level of density as nearby census block groups could unlock an additional 117,000 housing units in markets with medium-to-high housing demand, and earn nearly a billion dollars a year from the leases. They also point to precedents for this kind of thing, both in the US and abroad. What could be more in tune with abundance than a proposal that simultaneously eases housing shortages and enhances state capacity? — Matt Clancy
What is it like to know that a new drug exists – and has been proven safe and effective – but that it simply isn’t available where you live, and you might die without it? This was the reality for millions of Africans during the AIDS crisis in the 2000s. Even when tenofovir was approved in the US in 2001, most Africans couldn’t access it for nearly a decade, instead having to rely on a cheaper but toxic alternative that roughly a third of patients would abandon due to its side effects. These delays have shortened but remain prevalent across much of the world today.
The new magazine In Development has a great article on one solution: pooling regulatory reviews across countries, so that manufacturers don’t have to file separately in each one and overstretched agencies don’t duplicate each other’s work. The European Medicines Agency effectively does this across Europe, and the African Medicines Agency, launched in 2025 to do something similar across Africa. I learnt a lot from the piece about how it came about and the problems it faces today. Some of the continent’s biggest pharmaceutical markets haven’t signed on, some countries refuse to accept each other’s assessments, and the whole project runs on donor funding without a clear path to self-sufficiency. — Saloni DattaniBy now you’ve probably seen the viral clip of Marblehead resident David Modica asking his fellow townspeople the question heard ‘round the internet. Jerusalem Demsas wrote about it in The Argument, making the case that Marblehead’s decision to zone a golf course for housing (knowing nothing would ever get built) is what happens when local democracy mediates land-use decisions. She’s right.
But I sit on Marblehead’s Housing Committee, and I’ve been watching the aftermath up close. At our May 12 meeting, the mood, in part, was relief: the years-long 3A fight was finally over, compliance was in sight, and the town could stop bleeding resources on it. I get it. The town is in a budget crisis, and the committee chair’s own job has been on the line through all of this. Still, my reaction was that compliance isn’t the right metric for the Housing Committee, housing production is.
What strikes me, though, is that most of the people involved aren’t acting in bad faith, they’re operating within constraints they didn’t design. Voters first accepted (through record-breaking attendance at Town Meeting) and then rejected (through a dormant referendum process) the more meaningful plan. The state signed off on the weaker one. Everyone did their job, and the system still produced the wrong outcome.
We just started a team book club reading Alain Bertaud’s Order Without Design, and our first discussion kept circling this tension. Planners often want to do the right thing but are hemmed in by systems that make good outcomes structurally hard. Marblehead is a small-scale version of that story. — Nisha Austin
And a few grantee and team shoutouts and other announcements worth sharing:
Matt published “Social science at the NSF” earlier this week, looking at the history of fights over social science funding at the NSF and how the current proposed elimination of the SBE Directorate rhymes with the Reagan-era cuts of the early 1980s.
Progress Ireland published a series on housing in Europe, making the case that the EU has a significant opportunity to reshape the built environment in 2026 through permitting reform, the Construction Services Act, and Brussels-level policy change.
Mayor Mamdani released “Block by Block: The Housing Plan for a New Era,” New York City’s new 10-year housing plan, which aims to build 200,000 new affordable homes and preserve 200,000 more, backed by $22 billion in city housing funding over five years.