We’ve been writing regular round ups for a little while now, but so much has happened recently that this month’s post feels like it contains a year’s worth of breakthroughs. So pour yourself something cool, get cosy, and enjoy!

First, everything new in cancer. In our round up last month, we shared the news of daraxonrasib, the new breakthrough drug that roughly doubles survival in late-stage pancreatic cancer, which has long been considered untreatable.

This past week, however, there were results from many more major cancer trials presented at the American Society of Clinical Oncology conference (ASCO 2026), and we wanted to share some highlights.1

One presentation showed 7 year results for lorlatinib, a precision drug for metastatic lung cancers: specifically, a type (ALK-positive) that’s common in non-smokers. You rarely see a survival curve like it. 55 percent of patients were still progression-free after 7 years, versus just 3 percent on the older drug crizotinib.

A new prostate cancer drug, talazoparib, halved the risk of progression when added to hormone therapy, compared to hormone therapy alone, for prostate cancers carrying specific DNA repair gene mutations. Talazoparib belongs to a class of drugs (PARP inhibitors) that have already advanced treatment of breast and ovarian cancers with BRCA mutations. This trial suggests the same strategy also works for prostate cancer.

There was also long-term data on a recent endometrial cancer drug, dostarlimab, an antibody that was approved in 2023. The headline result after 4 years on the RUBY trial was that 58 percent of patients whose tumours carried ‘mismatch repair deficient’ mutations still hadn’t progressed when the drug was added to chemotherapy, versus just 16 percent on chemotherapy alone. That’s a huge difference! The drug is a ‘checkpoint inhibitor’: T cells have built-in ‘checkpoints’ that hold them back from attacking cells unnecessarily, but tumours exploit these to avoid being attacked, so blocking the checkpoint frees the T cells to attack the tumour.

Subscribe now

An early cancer detection blood test, the NHS Galleri blood test, was tested in a large RCT in England with about 142,000 people. The results are a little confusing, so let’s start by clarifying what a ‘good result’ here would look like: detecting non-benign cancers that would’ve missed otherwise, or catching them earlier so treatment could start sooner.

On the first count, it was successful: roughly quadrupling the overall detection rate and cutting cancers caught only at emergency presentation by around a quarter. On the second, it detected some cancers earlier: with a 16 percent rise at stage I–II, and 14 percent reduction at stage IV. But the trial missed its primary goal: there was no significant drop in combined stage III and IV cancers (because stage III diagnoses rose more than expected and roughly cancelled out the stage IV reduction). One way to interpret this – since the largest jump in detection was at stage III and the combined late-stage count barely changed – is that the ‘early’ detection test was mostly catching cancers at around stage III, rather than much earlier. That may still be useful, of course.

And finally, an mRNA cancer vaccine that might actually work. In people with high-risk melanoma, who’ve already undergone surgery, it halved the risk of cancer recurrence or death, compared to Keytruda (the blockbuster cancer immunotherapy drug) alone, in a phase 2b trial over 5 years.

Many previous mRNA cancer vaccines have been doomed by how they pick their targets. They sequence a tumour’s mutations and design the vaccine against the resulting ‘neoantigens’ (newly formed mutant proteins in a cancer cell), but most of those neoantigens are never actually displayed on the cell surface so the immune system can’t recognize them anyway. Melanoma is different: its neoantigens are readily displayed and recognize, which is why drugs like Keytruda already work well against it and why a vaccine has a much better chance of working as well.

Now, let’s move over to non-ASCO2026 news.

The most effective weight loss drug so far. Retatrutide, a new weight-loss drug, led to the largest reductions seen so far: roughly 28 percent loss in body weight at the highest dose in a phase 3 trial. The effect roughly matches the effects of bariatric surgery, where surgeons reduce the size of the stomach and reroute the small intestine so patients feel full earlier.

Retatrutide is an engineered peptide that acts on three receptors at once: the GLP-1 receptor, where it mimics a gut hormone that curbs appetite and slows gastric emptying; GIP, which also dampens appetite and improves how the body handles fats and sugars; and glucagon, which increases energy expenditure. It can do this because the three hormones are evolutionarily related, with similarly shaped receptors, and scientists engineered it as a single peptide that fits into all three of them.

Over the past few years, weight loss drugs have broadened in their effects: from one receptor (semaglutide) to two (tirzepatide) to three (retatrutide). More targets probably means higher efficacy in general, but it’s probably not linearly additive, with each added receptor contributing somewhat less than the last.

Subscribe now

The first in vivo gene editing therapy. A single injection has permanently edited away a rare disease: hereditary angioedema, a rare condition where a genetic mutation causes sudden, potentially fatal swelling attacks. In a phase 3 trial, a single dose of gene editing therapy reduced attack rates by 87 percent relative to placebo, and 60 percent of patients remained attack-free entirely over 6 months.

It’s notable because it’s the first time gene editing has worked inside a living human body (in vivo) in a late-stage trial. Previous gene editing medicines like Casgevy, the first approved CRISPR drug, for sickle cell disease, worked outside the body, ex vivo, meaning they involved removing cells from the patient, editing them in the lab, and replacing them. But this new gene editing therapy packages CRISPR machinery inside a lipid nanoparticle, a tiny fat bubble, that travels through the bloodstream to the liver and edits DNA there directly.

Editing within the body will make the therapy far simpler in the real world than editing outside the body. It’s likely easier to scale up, and opens the door to treating diseases where you can’t easily remove and return cells. For now, the lipid nanoparticle delivery system mostly targets diseases of the liver, but researchers are working to extend it to other tissues like the lungs and eyes.

Permanent gene-editing to reduce cholesterol. Speaking of gene editing: a gene therapy has now been used to permanently edit a gene in liver cells to lower LDL cholesterol. These results come from a phase I trial, in which 35 adults with an inherited disease causing very high blood cholesterol (hypercholesterolemia) or premature coronary artery disease got the therapy.

The therapy, called VERVE-102, is carried by a GalNAc lipid nanoparticle, which delivers it to the liver, where cholesterol is produced. There, the therapy edits the PCSK9 gene. PCSK9 regulates cholesterol levels in the body, typically reducing how much cholesterol is removed from the blood and taken up by liver cells. The way it does this is by binding to the LDL receptor, which grabs cholesterol, and tagging it for destruction, meaning less cholesterol is absorbed from the blood. So, by editing the PCSK9 gene, LDL receptors aren’t broken down as much, more cholesterol is taken up by liver cells from the bloodstream, and blood cholesterol levels are reduced.

At the highest dose, the average reductions were huge: PCSK9 levels fell by 88 percent and LDL cholesterol fell by 62 percent. These changes lasted more than a year, with no observed toxicity at the higher doses. VERVE-102 is possibly a one-off treatment, which is great because around 40 percent of people taking cholesterol-lowering medicines stop taking them regularly.

The target of the gene edit, PCSK9, has actually become one of the most heavily pursued ones in cholesterol medicine. Several recent drugs act on it, including injectable antibodies like evolocumab, which mop up PCSK9 directly, and the oral pill enlicitide. There’s also inclisiran, an siRNA drug that reduces the liver’s production of the protein, which is off-the-shelf and needs an injection just once every six months. The patients who’d benefit most from a one-time gene edit instead of those alternatives are likely not the wider population, but rather people with inherited conditions causing high blood cholesterol, who have a strong lifelong need to reduce their cholesterol and would most value a permanent fix.

An incredible feat of pharmaceutical synthesis. One of the recent drugs that helps patients lower their cholesterol level by blocking PCSK9 is called enlicitide. In a recent phase 3 trial, it lowered LDL cholesterol by an additional 50–60 percent for people already on statins.

Enlicitide is chemically fascinating for two reasons. One is that it’s formulated as a daily pill: all the other PCSK9 drugs on the market, including antibodies like evolocumab, the siRNA inclisiran, and the gene editor VERVE-102, have to be injected. The second is that enlicitide is a synthetic peptide. That’s surprising, because peptides are normally broken down in the gut before they can work. So how can a peptide be taken orally?

Well, enlicitide has an unusual structure. It’s a synthetic peptide made from eight amino acids that fold into a ring. This ‘closed’ shape helps the molecule survive the stomach and get absorbed in the small intestine. But the ring also makes enlicitide hard to manufacture. Some of its amino acids are ‘unnatural’, meaning they don’t appear in nature, so the peptide can’t easily be made by engineered cells, unlike other peptide drugs like semaglutide. And it’s hard to make chemically, too: the original synthesis took 63 reactions, and resulted in very low yields. The main bottleneck is the step where the chain closes into a ring, called macrocyclization.

Now, in a new study, Merck scientists improved the yield massively: with a 14-fold increase (to 70 g/L, at over 99 percent purity) by merging traditional chemistry with enzymes. They split enlicitide into three segments, called Northern, Eastern, and Western. The Eastern and Western parts were easy to make with conventional chemistry, but the Northern bit, which contains the unnatural amino acids, was instead built using enzymes. The team used directed evolution – repeatedly mutating an enzyme and keeping the best-performing ones over successive rounds – to engineer seven enzymes which, together with three natural ones, assemble the fragments and close the final ring. It’s one of the most ambitious feats of pharmaceutical chemistry ever published. And it shows that, even the most complex peptide drugs, made today at low yields, could in principle be manufactured at industrial scale through clever enzyme engineering.

Subscribe now

A cure for some patients with hepatitis B. Hepatitis B is a virus that causes liver damage and liver cancer in middle- and old-age. It’s preventable with birth-dose hepatitis B vaccines, but there are still around 250–300 million people globally who have chronic infections, mostly being infected before the vaccines were introduced.

Hepatitis B infections are particularly difficult to ‘cure’ because the virus has a little mini-chromosome that goes inside the nuclei of liver cells. So even if you destroy all the RNA transcripts, the mini-chromosome lives on and the infection persists.

But new drugs could destroy them. In two phase 3 trials, about 1,800 adults with chronic hepatitis B infections were treated weekly for 24 weeks with antisense oligonucleotides (short strings of nucleotides, called ASOs) or placebo. These ASOs bind to the virus’ RNA transcripts and mark them for destruction by enzymes. Then, the ASOs are recycled by the body.

Interestingly, even though the ASOs target RNA, they also reduce viral DNA, because hepatitis B can only make new DNA by reverse transcription (turning its RNA back into DNA), so cutting off the RNA cuts off the supply of new DNA. It also stops the virus churning out viral proteins like HBsAg, a surface antigen produced in enormous quantities to misdirect our immune cells.

About 20 percent of the patients who received the drug, called bepirovirsen, had a ‘functional cure,’ versus none in the placebo group. Here, a functional cure meant no detectable hepatitis B viral DNA and a loss of HBsAg sustained for at least 24 weeks post-treatment.

Human cells can ‘swap’ DNA with each other. It’s long been assumed that each human cell’s genome is sealed up in the nucleus and evolves independently of its neighbours. A new study finds that isn’t always true. The researchers found that large, chromosome-sized pieces of DNA can move directly from one human cell into another, lodge in the next cell’s genome, and remain functional, getting transcribed into RNA and translated into protein.

How on earth are they doing it? The authors observed DNA passing through ‘nanotubes’, which are thin bridges that briefly form when cells touch. Scientists have known about these for a few decades, and that cells were shuttling things like mitochondria through them, but it wasn’t confirmed that genomic DNA could be transferred through them until now.

To measure it, the researchers grew two batches of human cells together, and tagged DNA packaging proteins in each batch with a different fluorescent colour, one green and one red. They then stressed the cells to break their chromosomes or scatter them during cell division (using mitosis-blocking drugs, radiation, and CRISPR) and filmed what happened. On the time-lapses, they watched DNA of one colour travel down a nanotube into a cell of the other colour.

It’s not yet clear how much this matters in the body. But the authors note it could let genomically unstable cancer cells pass on their mutations to neighbouring cells. If that’s the case, the nanotubes themselves could be a target for cancer therapies.

CRISPR’s ancestors, hiding in bacteriophages. Researchers in Jennifer Doudna’s lab at UC Berkeley, which helped pioneer CRISPR as a gene-editing tool, have found a system that predates CRISPR and described it in two new preprints. CRISPR is a defense system bacteria use to recognize and cut up invading DNA, including from viruses. A subset of its proteins are built around a core structure called a RAMP, which has stayed remarkably constant over many millions of years, even as the genes encoding it have drifted apart. So the team used AlphaFold, the protein structure prediction tool, to hunt for proteins that look like RAMPs but are encoded by unrelated sequences. They found some hidden in bacteriophages, probably as a weapon that they use against rival phages.

They named the system VIPR, and it appears to be older than CRISPR itself. Like CRISPR, it’s guided by an RNA, but it differs in a few notable ways. It doesn’t cut DNA at all; instead it clamps onto a target sequence and silences the gene by physically blocking the enzymes that read it, leaving the DNA unaltered. It also finds that target with an unusual ‘noncontiguous’ code: the guide RNA reads the DNA two bases at a time and skips every third one, wrapping around the intact double helix rather than unzipping it. And it’s tiny: Vipr proteins are only about 180 amino acids, roughly an eighth the size of Cas9, paired with an RNA under 100 nucleotides. That means the whole thing can fit very easily into an AAV vector (often used as a delivery vehicle for gene therapy), and in principle you could pack many of them into a single therapy to fix several genetic errors at once.

A preventive antiviral for Covid-19, finally. The first oral drug to prevent COVID after exposure succeeded in phase 3 trials. It’s called ensitrelvir and is taken as a 5-day course of pills. In a phase 3 trial, people who took it after a household member caught COVID were 67 percent less likely to develop symptoms themselves.

It’s taken a long time to get here. A preventive antiviral pill is scientifically more challenging to develop than an mRNA vaccine, which mostly needs the specific genetic sequence (of the virus’s spike protein) to be identified and swapped in. In contrast, an antiviral like this is a custom-designed molecule that physically blocks a viral protein. That’s exactly why the vaccines were so important: they could be developed quickly, at a time when they were needed most.

Unlike the vaccines, ensitrelvir works by blocking the virus’s protease, a cutting enzyme it uses to mature its proteins. Because the protease mutates far more slowly than the spike protein the vaccines target, the drug should stay effective across naturally circulating variants. It could still face drug resistance over time, but the escape mutations seen so far seem to come at a cost to the virus’s own fitness. It’s worth noting that it’s approved in the US for post-exposure prevention specifically, not as a general pre-exposure prophylactic, and not for treating active infection.

Subscribe now

The first approved PROTAC drug. The FDA approved the first-ever ‘PROTAC’ drug, vepdegestrant, to treat a common form of advanced breast cancer. While most drugs work by blocking a disease-causing protein by binding to a pocket it uses to attach to its natural target, a PROTAC instead just destroys the protein. It’s a two-sided molecule that uses one end to grab the target protein and the other end to grab an enzyme (an E3 ubiquitin ligase) that tags the protein for recycling. This physically brings the two together, so the target protein gets marked for destruction and broken down.

Here, the target is the oestrogen receptor, which drives a common form of breast cancer. The drug treats tumours in which the receptor has mutated to stay permanently switched on and grows despite hormone therapy. In its phase 3 trial, it roughly doubled the time before the cancer progressed compared to the existing drug fulvestrant.

The approach could in principle reach the many disease-causing proteins that have no good pocket to block, which have long been considered ‘undruggable.’ This is just the first approved PROTAC, and many more are likely to come through the pipeline, and change the trajectory of diseases that we consider untreatable today.

What’s behind so much progress?

Part of it is timing: many cancer papers are published to coincide with the American Society for Clinical Oncology’s annual conference. But there’s more beneath that too.

Many new breakthroughs were precision drugs, targeted towards particular genetic mutations and tumours with certain genetic markers. Although the genomic revolution began a few decades ago, some of those advances have taken a long time to make it through the pipeline.

Another part is advances in structural chemistry. Daraxonrasib (the pancreatic cancer breakthrough), enlicitide (the oral PCSK9 cholesterol drug), and retatrutide (the new weight loss drug), for example, came from better understanding of the precise structure of proteins and how they interact with each other.

Both the genomic revolution and structural chemistry have helped researchers understand drug targets better and engineer them, including producing mutant KRAS proteins to find new druggable pockets for the pancreatic cancer drugs, and adapting new platforms, like the mRNA cancer vaccines, for melanoma.

This progress seems like an odd contrast with the cuts we’re seeing to science in the US, including to the National Cancer Institute, which is spending 80 percent less this year than its typical pace. It was responsible for funding fundamental research that eventually led to some of the breakthroughs in this post, including 1980s research into KRAS, which led to new pancreatic cancer drugs, and 2010s research into cancer immunotherapy targeted towards neoantigens, which led to cancer vaccines. The National Heart, Lung, and Blood Institute is also spending around 80 percent less than usual; decades ago, it funded fundamental research on LDL-receptor biology and PCSK9 that brought about cholesterol drugs.

In other words, the breakthroughs we’re seeing today are a lagging indicator of inputs into the pipeline: they reflect choices people made in the past. We’ll only be able to see the gaps created by reduced investment in retrospect, years from now, and see that technology isn’t the only reason for progress; sometimes it’s about the funding, institutions and incentives we create.

Illustration: Bacteriophages at work.

This is the last week to apply to Invisible College, our week-long seminar in Cambridge in August. If you are 18–22 years old, bright, ambitious, and interested in ideas: apply! Applications close on Friday 8th May.

The first gene therapy for deafness earns FDA approval. About sixty percent of all babies born deaf have some underlying genetic cause, and mutations in the OTOF gene account for between 2 and 8 percent of them. That gene encodes a protein called otoferlin, which enables hair cells inside the ear to transmit messages into the brain.

The OTOF gene is quite long, though, stretching about 90,000 bases of DNA. This means it’s too long to ‘package’ inside a single viral capsid for delivery. Regeneron, the company that developed the therapy, therefore decided to split up the gene, package each piece separately, and then deliver each half in its own viral capsid. The therapy was injected directly into the fluid-filled space of the inner ear, called the perilymph.

In one study with 12 children – all of whom were unable ‘to hear a gas-powered lawn mower’ before starting therapy – six could hear soft whispering without any hearing aids after 24 weeks of the therapy, and three others developed normal hearing. Besides being the first gene therapy for deafness, this is ‘also the first dual-AAV [adeno-associated viral capsid] therapy to be approved by the FDA’, writes Veera Rajagopal. The drug will be given away for free to patients in the United States, according to a Regeneron press release.

Does hair loss treatment actually work now? A new oral drug for hair loss actually seems to work, which feels surprising in a field that’s been plagued by scams and overblown claims for decades. Until now, there have been only a few approved treatments: oral finasteride (only for men, but coming with side effects like sexual dysfunction) and topical minoxidil (for men and women, applied to the scalp, with effects that are not long-lasting).

Now, Veradermics reports their drug, VDPHL01, grew around 30 additional hairs per square centimeter over six months compared to roughly 7 in the placebo group in a late-stage trial, and around 80 percent of participants reported improvement. The drug is a reformulated version of minoxidil engineered for slower, steadier release: standard oral minoxidil has a half-life of only a few hours; the idea now is to keep levels consistently high enough for continuous regrowth without reaching toxic levels. A second confirmatory trial in men is expected to report by the end of the year, which could result in FDA approval if it’s successful. There’s also an ongoing trial in women, where the options are even more limited.

Scientists watched a protein fold in real-time. A protein moves between its folded and unfolded forms in less than one-millionth of a second. For a new study, researchers captured this transition in real time for eight small proteins. They measured each protein’s folding by attaching red and green fluorescent dyes to amino acids that sit far apart when unfolded but come together when folded. As the dyes get closer, the green dye transfers energy to the red dye instead of emitting its own light. Whereas an unfolded protein will emit roughly equal numbers of green and red photons, a folded protein will emit mostly red photons. After watching these proteins fold, using high-powered microscopes, the scientists discovered (surprisingly) that large proteins fold faster than small ones. The smaller proteins had shorter waiting times – meaning they flip back and forth between states more frequently – but larger proteins had shorter transition times, meaning they completed folds faster once the process had begun. The largest protein transitioned in 0.7 microseconds, compared to 3.1 for the smallest. Evolution has probably optimized larger proteins to fold more efficiently via cooperativity, where one part of the protein coaxes another part to snap into place.

Subscribe now

Researchers recreate snake venom with organoids, to help research into anti-venoms. Snakebites kill 50,000 people each year in India alone. Computational biologists are using AI tools, such as RFdiffusion, to design proteins that can neutralize a few toxins in snake venom, and which don’t need refrigeration, but so far they only work against select proteins in the venom rather than the whole cocktail. To make antivenom today, scientists physically milk a snake by pressing on its venom glands so that it bites down and releases venom into a container. Then, they inject small, non-lethal doses of that venom into a horse, whose immune system produces the antibodies that are turned into the final antivenom. There are many ways to make this process far better!

In 2020, for example, Dutch researchers made organoids – spherical clusters of tissue grown from stem cells – for front-fanged snakes, including Elapidae and vipers. The nice part about these organoids is that they actually make venom that is chemically identical to the venoms made by living snakes. But for several years, nobody was able to make organoids for non-front-fanged snakes, which constitute about 70 percent of all snake species on Earth. These snakes have a different gland anatomy — they produce venom in a place called Duvernoy’s gland, which is smaller and harder to access than the venom glands of vipers and elapids, and extracting stem cells from it requires anesthetizing the snakes first. Recently, though, English scientists figured it out.

A new preprint describes how to make venom-producing organoids for snakes from the Colubridae family, which is the largest snake family on Earth and includes king, rat, and garter snakes. These organoids only make a tiny amount of venom (you’d need hundreds or thousands of them just to make enough to inject a single horse), but will be a useful starting point for basic research into antivenom design.

How long can you clone mice for? Scientists cloned a mouse for the first time in 1997 by taking a cell from an adult animal and injecting it into an egg lacking a nucleus. The egg, after being stimulated with chemicals, began to divide and gave rise to a mouse which was a genetic copy of the original animal.

Now the same research group have published a paper showing how they kept cloning the mice, repeatedly, for a period of 20 years. One might suspect that a lack of genetic diversity would make the animals non-viable (after all, that’s what happened to royal dynasties in which brothers married sisters or cousins). But instead, the researchers found that the cloning success rate stayed basically flat through 40 generations, but then started to fall.

The experiment ended after 58 generations, when the scientists could no longer clone the mouse. ‘By generation 57, the cloned mice had acquired over 3400 single-base changes relative to the starting mouse, whereas 62 generations of natural reproduction (in an inbred mouse strain) had accumulated 752’, writes Merrick Pierson Smela, a geneticist. ‘In other words, cloning caused 3.1 times more single-nucleotide mutations per generation than natural reproduction’. The mice also lost an entire X chromosome between generation 25 and 45, which was never regained.

Some ask ‘Where’s my flying car?’ but I ask ‘Where’s my omnibus p53 cancer drug?’ p53 is the body’s most important protein for protecting us against cancer. Roughly half of all solid cancers develop mutations in it that prevent tumor cell death, promote tumor growth, and prevent DNA damage repair. So why don’t we have effective drugs against it?

Unfortunately, there are thousands of different p53 mutations, and most work by changing the protein’s shape rather than knocking out a specific function, which makes it hard to fix with a drug. What’s worse is that mutant p53 is harder for cells to break down, so it builds up in high levels in cancer cells; drugs have a lot of dysfunctional protein to overcome.

Progress in p53 drug development seems to have been pretty hard. But some mutants are easier to target than others – like the Y220C mutant of p53, which is shaped with a pocket that a small molecule can slot into. That’s what some researchers have been targeting. A new phase 1 trial shows a proof-of-concept that this strategy can work using rezatapopt, a new compound that binds to the Y220C mutant p53 and reduces its levels in patients’ blood.

Although itʼs promising, this mutation is only carried by around 1 percent of p53 mutant tumors! And there’s still more to do to demonstrate its efficacy; various previous attempts at p53 drugs have failed in late-stage clinical trials. It’s a difficult problem, but one that would be hugely impactful if solved. Let’s hope this is a successful step towards it.

Subscribe now

Roughly a third of approved antibody drugs bind to proteins they’re not meant to. Monoclonal antibodies are used in medical treatment for various conditions including many cancers, and are generally described as extremely precise drugs that lock onto a single target – which is true, relative to other drugs – but it turns out that some also bind to unintended targets.

A new study screened 74 antibodies that were FDA-approved or in clinical studies against over 6,000 human proteins. The authors found that 28 percent bound to at least one unintended target, which could cause unexpected side effects or reduce the efficacy of the drugs, although that wasn’t tested directly. On the bright side, the authors showed that inducing just one or two mutations to the antibody were enough to remove these off-target bindings without touching the intended ones. This kind of early systematic screening isn’t standard practice yet, but it could probably help find better antibody candidates and reduce failure rates in clinical trials.

US science agencies have money but aren’t spending it. A new analysis tracking real-time spending data finds that NSF is issuing new grants roughly 70 percent slower than its historical pace, NIH about 50 percent slower than before, and the National Cancer Institute particularly bad at 79 percent slower. While Congress actually held firm on science budgets, the current administration has been slow-walking the actual disbursement of funds to researchers, as it did in FY2025 before advocacy pressure forced a correction.

To make it easier to understand these slowdowns, Jordan Dworkin, Saloni’s colleague at Coefficient Giving, built sciencespending.org, a public tracker of grant-making pace across five major agencies. The problem is worse than just delays, because unspent NIH funds expire at the end of the fiscal year and go back to the Treasury, which means that money may never reach researchers.

Eugene Braunwald, probably the most important figure in modern cardiology, died on April 22nd at age 96. It’s hard to overstate his influence. Before his work in the 1970s, a heart attack was treated with sedation and bed rest: the prevailing view was that once muscle was dying, little could be done. Braunwald overturned this by showing that the amount of damage from a heart attack depended on the balance between oxygen supply and demand in the heart, and that one could act fast to shift this balance and rescue threatened muscle. The idea became known as ‘time is muscle’ and underpins almost all modern reperfusion therapy: clot-busting drugs, angioplasty, and stenting, which help clear arteries and deliver oxygen to the heart quickly.

He also defined the mechanics of hypertrophic cardiomyopathy (a condition where the heart muscle thickens and obstructs blood flow), and in 1984, he founded the TIMI study group, which has since run over 70 randomized trials shaping standard care for heart attacks, heart failure, and diabetes.

Over the course of his career, the share of patients who died within a month after having the most severe type of heart attack fell roughly three-fold – from a 30-day mortality rate of around 17 percent to around 6 percent; much of that is attributable to treatments based on his research.

It’s striking how recent all of this is: the idea that a heart attack was something you could actively treat is only about 50 years old! (Part of the reason is that people hadnʼt invented ways to study the mechanics of the living heart without killing the patient until the mid-20th century.)

Organ transplants have become much more efficient. Most organ transplants in the US come from deceased donors, and technology has made it much easier to recover those organs and keep them alive for transplants. A new study, looking at all organ transplants from deceased donors, finds a massive rise in the number of organs recovered for donation in the last five years. In 2000, around 2 percent of donors who died from circulatory death were able to donate their organs; that figure rose to 49 percent in 2025.

The authors attribute this to the use of normothermic regional perfusion, a procedure that’s used to restore blood flow to organs after they’re removed from deceased donors, which became more commonly used since 2019, and new machines to perform this procedure for the liver, which were approved in 2021.

Semaglutide doesn’t slow down Alzheimer’s disease. GLP1 drugs like semaglutide have shown benefits for weight loss, diabetes, kidney disease, and cardiovascular disease, and observational data suggested they might also reduce dementia risk. But two large phase three trials of nearly 4,000 patients with early confirmed Alzheimer’s disease have found no effect on disease progression across cognitive and functional measures.

[Saloni: I expected a slight benefit of GLP-1 drugs through their cardiovascular effects in reducing vascular dementia, which can co-occur with Alzheimer’s, though not a direct benefit.] It’s a useful reminder of the importance of randomized controlled trials, even when there’s plausible evidence or observational data supporting an effect. In this case, it was likely the result of confounding. Sadly, the list of failed Alzheimer’s drugs is long.

A transformative new drug for pancreatic cancer. Pancreatic adenocarcinoma is one of the deadliest cancers, with a five-year survival rate of just 13 percent; advances in therapy against the disease have been slow. A key reason is that around three-quarters of these cancers are driven by mutations in KRAS, a protein that has been considered essentially undruggable because it lacks an obvious binding site for a drug molecule. A new treatment, called daraxonrasib, gets around this by first binding to another protein in the cell, cyclophilin A, forming a complex that can then latch onto and block mutant KRAS, cutting off the signal that drives tumor growth.

In a phase three trial of around 500 patients with metastatic pancreatic cancer who had already progressed while on chemotherapy, the company reported a median overall survival of 13.2 months with daraxonrasib versus 6.7 months with chemotherapy, roughly doubling survival.

That is an unprecedented result for a cancer that has so far been devastating and stubbornly resistant to progress. These are topline numbers released by the company, and detailed data on safety and subgroup analyses will help understand the result better. The drug also appears to cause significant side effects, including serious bleeding and skin problems. But it may also be one of the biggest cancer breakthroughs of the year, if not decade.

This trial enrolled patients with late-stage pancreatic cancer, but phase three trials are also underway testing the drug as a first-line treatment and as an add-on for patients whose tumors are still surgically treatable. The drug is also being tested in lung cancer, and could be expanded further; KRAS mutations drive roughly 20 percent of all human cancers. Since a viable mechanism for targeting KRAS has been demonstrated to work in patients, researchers can focus on iterating, improving safety, optimizing dosing, testing combinations, pushing the drug earlier in the disease course, and expanding its use for other cancers.

Weight loss drugs that also preserve muscle mass are moving through the pipeline. A well-known side effect of GLP-1 drugs like semaglutide is that around a quarter of the weight lost comes from lean tissue rather than fat, because the drugs work by suppressing appetite rather than directly targeting fat.

A new phase two trial tests whether combining semaglutide with bimagrumab, a new antibody, can address this. Bimagrumab blocks activin receptors, which normally promote fat storage and limit muscle growth, and the idea is that the two drugs work through distinct mechanisms, so their effects on fat loss should be additive, while its muscle-preserving action could offset muscle loss from semaglutide. The trial found that people lost much more fat on the combination treatment than with semaglutide alone, 34 percent compared with a 21 percent reduction in fat mass, while they retained most lean mass, with a 1–2 percent reduction compared to a 7 percent in the semaglutide-only group.

It’s worth noting that bimagrumab has its own drawbacks: it’s given by intravenous injection, which will likely reduce its usage, and likely caused muscle spasms and acne in roughly half the patients in the trial.

Subscribe now

Cell simulation caveats. Researchers built a computational model that simulates one cell division, about 100 minutes of biological time, for a single bacterial cell. Each simulation takes four to six days to run on two NVIDIA A100 GPUs and includes some randomness, meaning each simulation plays out slightly differently. After running the simulation 50 times and averaging the results, researchers found they could make a few predictions about how cells work without fitting to experimental data. For example, the simulated cells divided every 105 minutes, which matched experiments, and mRNA molecules had an average half-life of 3.63 minutes, which is close to ground truth.

The cell they’re modeling is JCVI-syn3A, a highly-engineered bacterium with a minimal genome of just 493 genes, compared to over 4,000 for E. coli. Its proteome, transcriptome, and metabolism have all been studied in depth, making it a really good starting point for whole-cell simulation. Rather than simulate every molecule individually (which would vastly exceed the capabilities of any modern computer), the authors instead wrote down all the stuff that happens inside a cell, like transcription, translation, and metabolism, and decided which type of mathematical model would be best-suited to describe each thing. Some cell processes were modeled deterministically, others had ‘spatial’ elements, and other parts were relatively random.

The final simulation had four parts: A Reaction-Diffusion Master Equation (RDME), used to model the individual proteins, RNAs, and ribosomes; a Chemical Master Equation, used to model things where spatial location doesn’t matter as much; Ordinary Differential Equations, used to model changes in metabolite concentrations, and Brownian Dynamics, which simulated the chromosome as a physical chain of beads.

All four models are run together, using a script that synchronizes their results with each other. The Reaction-Diffusion equation takes time steps of 50 microseconds of biological time. Every 12.5 milliseconds of biological time – or after the Reaction-Diffusion equation runs 250 times – the simulation pauses so that the other models can synchronize based on the latest state of the simulation.

This simulation seems to be a major advance over any other existing whole-cell models, but it still does not include a lot of things. For example, the simulation doesn’t account for polysomes, which are clusters of ribosomes that all latch onto a single mRNA and translate at the same time. Polysomes are really common inside of cells, but this simulation assumes that each mRNA can only be translated by one ribosome at a time. Also, it does not include polycistronic transcription. In bacteria, genes are often grouped next to each other on the chromosome and thus transcribed – turned into mRNA – all at once, together. The majority of genes in E. coli, for example, are arranged in these operons, and the authors of this paper acknowledge that many syn3A genes are likely co-transcribed the same way. But the simulation doesn’t capture it.

With that said, this paper is important because by trying to simulate a cell, we are able to figure out what we don’t yet understand and, therefore, which experiments we ought to perform to reconcile those gaps.

A highly effective new hypertension drug. The body regulates blood pressure partly through aldosterone, a hormone that tells the kidneys to retain salt and water. Some people with hypertension have very high aldosterone levels.

A new drug, baxdrostat, blocks the enzyme that produces aldosterone, reducing how much salt and fluid the kidneys retain. This phase three trial tested the drug in patients with ‘resistant hypertension’, who were already on ≥3 blood pressure drugs but still couldn’t get their readings within recommended levels, <140 mmHg. Researchers measured their ambulatory blood pressure with a monitor worn for 24 hours. That’s much better than single clinic readings, which can be inflated by anxiety and miss overnight blood pressure, which is a strong predictor of heart disease risk.

After 12 weeks, baxdrostat reduced systolic pressure by about 14 mmHg compared to placebo; that makes it one of the largest reductions seen with an antihypertensive drug.

Subscribe now

The brain shields tumors from the immune system. For a new study, researchers mapped how a lung tumor talks with the brain in mice. Lung tumors secrete proteins that attract sensory nerves from the vagus, which connects internal organs to the brain. Some of these neurons, which are marked by a specific protein called NPY2R, then send signals from the tumor through the brainstem. In response, the brainstem ramps up sympathetic nerve firing, coaxing noradrenaline to go into the tumor and shield it from attacks by the immune system. When researchers destroyed this circuit, either by silencing the sensory neurons with chemicals or deleting the proteins involved entirely, the lung tumors grew more slowly, the immune system, specifically T cells, was more active, and the mice lived much longer. The next step will be to figure out whether similar circuits exist for other cancers and in other organisms, including humans, before building medicines to disrupt it.

We finally know how plants make nicotine. Nicotine, the molecule, was first isolated nearly 200 years ago by German chemists. It is made from two ring-shaped molecules fused together, but nobody quite knew the enzymes responsible for that final step; the fusion. A new study in Cell, from a group in Shanghai, identifies the enzymes responsible and maps the full pathway for the first time.

An enzyme called A622 strips carbon dioxide from one of the rings, called nicotinic acid, thus making the molecule highly reactive. This reactive intermediate then attacks the other ring, joining the two halves together. Other enzymes strip off a lingering sugar group to make nicotine. Figuring this out took many years of work, as the scientists grew 643 inbred plant lines by crossing together 26 different parent tobacco plants. They finally stumbled upon a mutant plant which was not able to make nicotine, and then sequenced its entire genome. They also crossed back this plant and inbred it for two generations to find the mutation responsible; a single C-to-T swap. With these data and other metabolism experiments (in which plants are fed with heavy isotopes, and those atoms get tracked through the plant), they identified the missing enzymes and biosynthetic steps. And then, just for good measure, the researchers rebuilt the entire pathway in yeast, tomato plants, eggplants, and peas using synthetic genes, just to prove it all worked as they thought.

Did AI cure a dog of cancer? The most viral health story of late concerns Paul Conyngham, a Sydney tech entrepreneur with no biology background who used AI tools to design a personalized mRNA cancer vaccine for his dog, Rosie. Conyngham paid $3,000 to sequence Rosie’s tumor DNA, used AI tools to study the sequence, and then worked with a nearby university to manufacture and inject a custom mRNA vaccine. ‘The tennis ball-sized tumour on Rosie’s hock has shrunk in half’ as a result, according to reporting in The Australian.

Making a one-off mRNA vaccine is not especially hard, as bioengineer Patrick Heizer explains. Researchers cure mice of cancer routinely. It’s much harder to prove that a therapy is both safe and effective in human trials. The $3,000 figure mentioned above is also only the cost of the sequencing itself. The actual cost for these experiments, including reagents and labor, was likely far higher.

Palli Thordarson, Director of the RNA Institute where the work was done, further clarified that not all of Rosie’s tumors responded to the vaccine. The team is investigating whether the dog’s various tumors each mutated in unique ways, which may have made the vaccine less effective against some of them. Thordarson also said the treatment involved the co-administration of a checkpoint inhibitor, anti-PD-1, which is itself a highly effective anticancer drug. This confounds the results, making it unclear whether the mRNA therapy or the anti-PD-1 drug was most useful for Rosie.

This story is most useful, in our eyes, for what it says about the ridiculous, bureaucratic hurdles required for clinical testing, even on dogs. ‘The red tape was actually harder than the vaccine creation, and I was trying to get an Australian ethics approval and run a dog trial on Rosie,’ said Conyngham. ‘It took me three months, putting two hours aside every single night, just typing the 100 page document.’ Similar bureaucracy pervades clinical trials in the US and UK, and we should urgently find faster ways to design and test drugs in human patients.

A new way to hide transplanted cells from the immune system

Recipients of organ transplants are typically given lifelong immune suppression after the procedure to prevent their immune cells from attacking donor cells.

But this leaves recipients chronically vulnerable to infections, and so for a long time, researchers have looked for ways to avoid suppression. Methods have included encapsulating donor cells within membranes, using donor bone marrow to retrain the immune system, and delivering immune-modulating drugs only to the graft site.

These methods are good, but imprecise, as they rely on adjusting the immune system broadly. A new approach instead focuses on the transplant cells themselves: a 42-year-old man with type 1 diabetes has become the first person to receive a transplant with gene-edited cells that are less noticeable to the immune system.

In this case, the immune protection is built into the donor cells themselves, potentially allowing doctors to forgo the need to match donors entirely.

The scientists used CRISPR-Cas12b and a lentivirus vector to inactivate two genes in the pancreatic islet cells that were being transplanted. Inactivating the genes removed class I and class II HLA molecules, which are tags that immune cells use to recognise foreign tissue. The scientists also added a CD47 protein, which signals ‘don’t eat me’ to immune cells.

In the first 12 weeks after the transplant, the man’s immune system responded to unedited cells that remained in the graft, but showed no reaction to the fully edited cells. The edited cells survived and produced insulin, without causing any notable side effects.

In principle, the same strategy could be applied to many donor-derived cell therapies: from heart muscle cells for repairing damage after a heart attack, to neurons for Parkinson’s disease, to retinal cells for restoring vision.

Subscribe now

AI tools create bespoke antibiotics

In the last few years, there have been many papers that use AI tools to either design or discover new antibiotic molecules. Perhaps most famously, in May 2023, researchers used a neural network to discover abaucin by testing 7,500 molecules against A. baumannii and then used those data to train the neural network to predict whether other molecules would also be effective against the microbe.

The abaucin paper proved that AI could find hidden antibiotics amongst existing chemicals; a new study by Aarti Krishnan and colleagues shows that AI can create entirely new antibiotics; chemicals that do not exist in nature.

The researchers began with an AI model that had previously been trained on ‘growth inhibition data for 38,765 compounds screened against’ two microbes. Next, they built a database of >45 million chemical fragments ‘with chemically meaningful structures’, and automatically compared them to a ‘manually compiled set of 559 known antibacterial compounds’. They did this step to make sure that their generated batch of 45 million chemical fragments would sample a wide swath of potential chemicals not covered by existing molecules.

Then, the authors screened this large fragment library using an AI model called Chemprop which predicts whether each molecule is likely to have antibiotic activity. More than 1 million molecules passed this filter. After filtering out molecules likely to be toxic, and doing some further computational checks, the researchers ultimately selected 80 compounds that were ‘structurally dissimilar from each other and had the highest predicted antibacterial scores’. Of those 80 compounds, two could be chemically synthesized with high purity. Those two molecules were called NG1 and NG2.

NG1, but not NG2, was then shown to inhibit growth of multidrug-resistant N. gonorrhoeae in the laboratory. This AI-designed antibiotic is structurally distinct from all known antibiotics and, according to the authors, likely acts through an entirely uncharacterized mechanism to kill microbes.

Paper mills are spreading

There is a large industry in science where companies known as ‘paper mills’ sell fake papers. For a fee, these firms produce bogus research results and offer them to scientists who need publications to get jobs, grants, or promotions. According to a new paper, there are now brokers, middlemen, editors, and even entire journals cooperating with these paper mills.

To figure this out, the study authors turned to an open-access journal, called PLoS One, which publicly discloses which editor accepted each published study. (This transparency is a good thing, and the journal was only used as a case study.) The study’s authors used statistical tests to check whether specific editors at PLOS One were likely to be working with paper mills.

Between 2006 and 2025, the journal relied on 18,329 editors to publish more than 276,000 articles, of which 702 were ultimately retracted. Using a Poisson binomial test, the study’s authors calculated ‘whether each editor accepted ultimately retracted...articles significantly more often than expected by chance alone.’ The answer was yes: 22 editors at PLoS One accepted articles ‘that were retracted significantly more frequently than one would expect by chance.’

These editors oversaw just 1.3 percent of all articles published in the journal, and yet those articles accounted for 30.2 percent of all retractions. The problems have been worsening, with the number of papers suspected to have been produced by paper mills doubling every 1.5 years even though total publications have only doubled every 15.

Subscribe now

Clinical trials

• A new long-acting drug could change how we prevent the flu. After a single injection lasting six months, Cidara’s experimental antiviral reduced flu by around 76 percent in a phase 2 trial of 5,000 adults – a performance comparable to the most effective flu vaccines. Antivirals with such long-lasting effects are becoming more common as researchers make progress on ways to release drugs more slowly within the body, and identify drugs that remain potent at very low doses. Cidara’s next trials will test whether combining the antiviral with vaccines could offer even stronger protection for vulnerable groups.

• A new early-stage trial suggests that mRNA technology could be used to overcome some of the problems with HIV vaccines. Delivering the HIV’s envelope protein as a vaccine changes its shape. This exposes irrelevant regions the immune system can easily target, and distracts antibodies from the hidden, protective sites that matter most. The new mRNA vaccines avoid this problem by anchoring the envelope protein in the membranes of cells, preserving its natural shape. In 108 healthy volunteers, about 80 percent of those who received the membrane-bound versions made antibodies that blocked infection in lab tests, compared to only 4 percent with the soluble-protein version. The results are promising but still preliminary. The long-term outlook, however, is uncertain, as US government funding cuts to mRNA vaccine research will slow or stall progress in this area.

Lab science

• Most people think of CRISPR proteins such as Cas9 as DNA editors, but a new paper shows it is relatively straightforward to reprogram these proteins to target RNA instead. Deleting a stretch of just 55 amino acids from IscB (the evolutionary ancestor of Cas9), for example, made the protein lose its ability to bind DNA and instead prefer interactions with RNA. The truncated protein could cut mRNAs and silence gene expression. The same strategy also worked on other Cas proteins, including Cas9, suggesting many DNA editors are perhaps more versatile – and could therefore be reprogrammed to fix more diseases – than once thought.

• Progeria, a rare genetic disease, affects about 150 children worldwide. It is caused by a single nucleotide mutation in the genome and causes rapid aging. Fortunately, there has been major progress in treatment in just the last two decades. The first FDA approved drug, lonafarnib, extended life expectancy by 2.5 years. But a different approach, which relies on base-editing therapies that use a modified form of gene-editing tools to ‘swap’ out one nucleotide in DNA for another, has now cut toxic progerin protein levels by 90 percent and tripled survival in mice. Human trials are expected soon.

• Scientists have reconstructed the first full genome of the 1918 pandemic flu virus. It came from an 18-year old man who died in Switzerland from the first wave of the pandemic and whose sample was preserved as a specimen at the University of Zurich’s Human Remains Collection. To reconstruct the viral genome, they developed a new method for sequencing ancient RNA that works well on tiny, degraded fragments (the only kind left in such old tissue samples). Remarkably, the virus already carried genetic changes linked to adaptation to humans, meaning that these mutations had arisen very early in the pandemic.

The second patient in the world to receive ‘prime editing’ gene therapy for rare immune disease

A 56-year-old man with chronic granulomatous disease (CGD), which causes repeated infections and chronic inflammation of the intestine, became the second person in history to be treated with prime editing, a gene-editing technique that can rewrite DNA with high precision. His form of CGD was caused by a missing ‘GT’ in the NCF1 gene, which prevents immune cells from producing an enzyme called NADPH oxidase.

Prime editing is part of the CRISPR toolkit, but it adds new features that make it more precise and versatile.

Standard CRISPR-Cas9 works like molecular scissors: it cuts through both strands of DNA at a target site, and the cell’s own repair machinery fills in the gap. The process can be unpredictable.

Prime editing, by contrast, makes only a single-strand cut and brings along its own repair kit (which includes a modified Cas9 protein to find the target, a guide RNA to position it, and an attached reverse transcriptase enzyme that writes in the new, correct DNA sequence directly from a built-in template). This lets scientists change DNA bases exactly where needed, without relying on the cell to guess how to fix the break.

This makes it ideal for CGD, which is caused by tiny but critical errors, such as the patient’s missing ‘GT’ in the NCF1 gene. In this case, doctors removed his bone marrow stem cells, used prime editing to insert the missing letters, and returned the corrected cells to his body. Just 15 days later, the edited cells had been engrafted, and his immune cells regained enzyme function. His intestinal inflammation resolved.

David Liu, who pioneered this gene editing technique, explains how it works in a video below.

However, Prime Medicine, the biotech company that developed the therapy, has announced it will not continue developing this particular therapy. For ultra-rare diseases like this, the number of potential patients is tiny, and manufacturing each dose is complex because it’s individualized. Prime Medicine is looking for external partners to carry it forward.

A high-resolution, 3D structure of the entire E. coli genome

Genomes are often taught in school as being one-dimensional objects. A big enzyme called DNA polymerase zips along the DNA strand and makes copies prior to cell division. Another enzyme, called RNA polymerase, converts DNA into RNA in preparation for protein synthesis. And so on.

In reality, however, genomes are not one-dimensional objects defined solely by the sequence of their letters. Their shape is quickly changing, both in time and in space.

A new study in Nature reports this three-dimensional dynamism for the entire E. coli genome at a high resolution. With this map in-hand, researchers can clearly delineate which parts of the genome are physically close to each other in the cell, even though they may be far apart along their DNA sequence. One major finding is that clusters of related genes often fold into loops, bringing their start and end close together. This likely makes transcription more efficient, since enzymes can quickly loop back to the beginning instead of diffusing away.

The researchers created this map using a method called Micro-C. First, the cells were treated with a crosslinking agent that glued together any proteins and DNA that were physically touching in the cell. Next, researchers added an enzyme, called a nuclease, that chopped up the genome into tiny pieces. Last, an enzyme called DNA ligase was used to join tethered fragments together into a single molecule, which was then sequenced on a normal DNA sequencing machine. A computer algorithm then looked at which base pairs were joined together, and used that information to re-create a 3D map.

Subscribe now

Clinical trials

• A newly tested parasitic drug, moxidectin, looks like it could help tackle whipworm. Whipworm infects hundreds of millions of children in poorer rural areas, stunting growth and learning, and has little response from standard deworming pills, like albendazole. In a phase three trial in Tanzania, a single combination dose of moxidectin plus albendazole cured 69 percent of infections (meaning infected kids no longer had detectable worm eggs in their stool two to three weeks later) compared to just 16 percent with albendazole alone.

  

• Treating narcolepsy is hard: it is difficult to make a pill that can reach the brain, activate orexin receptors strongly enough to promote wakefulness, and still be safe. But a pharma company, Takeda, has reported that a new drug, ‘oveporexton’, has met its targets in phase 3 trials, making it likely to be approved by the FDA. In two studies involving 270 patients with narcolepsy type 1, subjects taking the drug had close to normal wakefulness in standardized lab tests, and reported less daytime sleepiness and improvements in their attention, daily functioning, and quality of life, compared to those on the placebo.

  Subscribe now

Lab science

• Scientists have determined how antivirals could shut down the measles virus’s replication machinery. Measles still causes more than 10 million infections and over 100,000 deaths a year, mostly in unvaccinated children, and there are no approved antiviral drugs to help prevent or treat it. Using a technique called cryo-electron microscopy, they showed that a small molecule, named ERDRP-0519, latches onto a key control loop inside the virus’s polymerase enzyme, locking it into an ‘off’ position so it can’t copy RNA. The same drug also binds to the polymerase enzyme of Nipah virus, a highly fatal relative of measles. This suggests a single compound could target two of the most dangerous viruses in the Paramyxoviridae family.

  

• Protein nanoparticles are tiny self-assembling structures that researchers often use as scaffolds to display vaccine antigens, hold enzymes together, or boost cell signaling. Historically, scientists have designed new types of protein nanoparticles which are symmetrical, meaning every side looks the same. Symmetry makes these nanoparticles easier to build, but it also limits what they can do. But now, researchers at the Institute for Protein Design have designed a new kind of protein nanoparticle with two distinct sides, meaning the same nanoparticle can interact with multiple different molecules at once. The researchers first built two halves of the nanoparticle, each with a slightly different set of subunits, and then joined them together. To demonstrate it worked, they engineered the particle’s structure so that each face displayed additional binding sites (‘minibinders’) pointing outward. The finished particles could then pull together two different types of receptor-coated beads. The approach could open the door to more versatile vaccines, diagnostics, and therapies.

• It may now be possible to make CAR-T cells – a powerful form of immunotherapy used to treat cancer and certain autoimmune disease – directly inside the body using nanoparticles, instead of having to manufacture them in a lab. That matters because today’s CAR-T treatments, which reprogram a patient’s own T cells to hunt disease-causing cells, are life-saving but costly, slow, and out of reach for most patients. Working in mice and monkeys, scientists built lipid nanoparticles that carried mRNA instructions for a cancer-fighting receptor and targeted them to T cells. The treated animals rapidly generated their own CAR-T cells, which eliminated B cells. The next step would be to show that this translates to clearing B cell tumours or blocking the progression of autoimmune disease.

• A new blog from OpenAI and Retro Biosciences describes a custom AI model, ‘GPT-4b micro’, that can design more efficient Yamanaka factors, or the proteins used to ‘reprogram’ mature cells back into stem cells. Shinya Yamanaka first discovered these proteins in 2006 by narrowing down 24 candidates down to just four essential ones: OCT4, SOX2, KLF4, and c-MYC. The new AI model (trained on an unreleased dataset incorporating protein sequences, biological text, and 3D structures) ‘designed’ new versions of SOX2 and KL4 that, when introduced into fibroblast cells, improved reprogramming efficiency by about 50-fold compared to natural forms of the proteins.

• A large study, tracking the careers of more than 12,000 US faculty across 15 disciplines, measures just how much getting tenure (or a permanent faculty position) shapes research productivity. Publication rates rise steeply during the tenure track and peak just before tenure. In lab-based fields (like biology and chemistry), research output stays high afterwards and tenured faculty keep publishing at basically the same pace. But in non-lab fields like mathematics and sociology, productivity falls. After tenure, faculty are more likely to pursue novel or higher-risk projects, but their work garners fewer highly cited ‘hit’ papers.

What we enjoyed reading

• AI will not lead to an Alzheimer’s cure. But will it help? In a mini series of blogposts, Jacob Trefethen describes ten bottlenecks in medical research, and explains compellingly where AI can make the most impact and why it won’t be enough for medical progress.

• The FDA has a huge (yet inaccessible) archive of drug submissions. These files include thousands of pages covering trial design, manufacturing, safety, and regulatory negotiations. And many of the companies that submitted this paperwork have gone under! In a report for the Institute for Progress, Ruxandra Tesloianu explains how, by using bankruptcy law to acquire and publicly publish FDA filings from failed drug sponsors, the US could democratize regulatory know-how, and perhaps power AI systems that level FDA filings for smaller biotech companies.

• An excellent post by Julia Rohrer about how correlations are more complicated than people think they are, how to interpret them properly, and how ‘things can get really messy once you calculate multiple correlations and compare them.’

Doctors made a custom CRISPR therapy to treat an infant with a rare genetic disease

A baby, named KJ, was born with a deficiency in carbamoyl phosphate synthetase 1 (CPS1). This deficiency means his liver couldn’t convert ammonia – which is made naturally when the body breaks down proteins – into urea, damaging his brain and liver. Babies born with CPS1 deficiency often end up staying in the hospital until they’re eligible for a liver transplant. They have to follow strict diets; specifically, a mix of natural protein and a medical formula that supplies only essential amino acids. If a baby’s protein intake is too high, their ammonia levels can spike, putting them at risk of permanent brain damage or death.

Researchers at the Children’s Hospital of Philadelphia diagnosed KJ within days of birth, and then spent about six months designing and testing a personalized gene-editing therapy to fix his deficiency. They used a CRISPR ‘base editor’ – a tool that rewrites a single DNA letter without breaking the DNA – delivered to his liver using lipid nanoparticles. KJ received the first dose in February this year. By April, he had received three in total.

Thanks to the therapy, KJ began tolerating more protein in his diet, and his blood ammonia levels dropped. He was able to reduce medications, gain weight, and move up on the growth chart. Because KJ’s condition was life threatening and rare, the team was able to administer the therapy under the FDA’s ‘single-patient expanded access’ pathway. Instead of going through the years-long clinical trial process, KJ’s treatment received FDA approval within weeks. This approach – sometimes called ‘N-of-1’ treatment – shows how regulatory systems can be flexible in urgent cases and might pave the way for faster development of personalized gene therapies in the future.

Subscribe now

Arc Institute discovery allows longer gene edits

CRISPR tools have a shared limitation: they can only make small edits to genes, usually just a few letters long, which are helpful only for diseases caused by small, specific mutations, like KJ’s. In contrast, bridge recombinases, a new gene editing technology, are designed to work on a much larger scale. Like CRISPR, they use a programmable RNA molecule to find their targets, but instead of one RNA guide, they use a special ‘bridge RNA’ that folds into two loops. One loop binds to the target DNA, the other to the donor DNA, bridging them together. This allows the enzyme to rearrange large segments of DNA precisely.

In 2024, researchers at the Arc Institute discovered this mechanism and then engineered a version, called IS622, that works in human cells. They showed that it could make edits of nearly a million base pairs in length. The median human gene is around 24,000 base pairs in length, so a million base pair edit could affect the equivalent of around 40 genes. (The largest known human gene – dystrophin – is just over 2 million base pairs long.)

This means that bridge recombinases could allow scientists to edit much larger regions of the genome – including entire genes, long regulatory elements, or clustered gene families – which are out of reach for CRISPR-based tools.

New RSV vaccines are reducing infant hospitalizations in the US

Respiratory syncytial virus (RSV) is the leading cause of hospitalizations in infants in the US. Recently, however, scientists succeeded in developing two powerful tools for protection: a maternal vaccine that passes on protection to newborns and a protective antibody called ‘nirsevimab’, which is given directly to babies. Saloni previously wrote about RSV vaccines back in 2022. Since then, large phase three randomized trials have shown that each can reduce infant hospitalization rates by 80–90 percent. They were approved in mid-2023 and rolled out widely for the 2024–25 winter season.

Among the youngest infants (0–7 months) hospital admissions for RSV were around half the previous norm, as the top panel of the chart shows. Around half of infants this age received the vaccine or antibody. By contrast, hospitalizations for older toddlers (in the lower panels), most of whom didn’t receive the new products, stayed in line with or above past seasons. Widespread use of the new RSV vaccines and antibodies is already translating into fewer hospital beds filled with babies. Remember that only around half of the eligible infants received the vaccines that season. If vaccination rates were higher, the reduction in hospitalization rates would be even greater.

A new cholesterol drug

Lowering low-density lipoprotein (LDL) – the particle that carries cholesterol into artery walls – helps reduce the risk of heart attacks and strokes, and statins have helped millions do just that. But even the highest doses cannot reduce many high-risk people to the LDL targets that modern guidelines recommend.

Obicetrapib is a new drug that may help close that gap. It belongs to a drug class called CETP inhibitors, which block a protein that moves cholesterol between lipoproteins in the blood. Earlier CETP inhibitors had disappointing results due to side effects. But obicetrapib is designed to bind CETP more precisely, with fewer side effects.

Obicetrapib cut LDL cholesterol levels 30 percent in a phase three trial involving over 2,500 people with a history of heart disease or inherited high cholesterol. All were already on the maximum doses of statins and other lipid-lowering drugs, and with LDL levels above the target level. Importantly, the rate of side effects was similar between the two groups, suggesting the drug was tolerated well.

Subscribe now

Silencing lipoprotein(a)

Another source of cardiovascular risk is lipoprotein(a), or Lp(a). Lipoprotein(a) is similar to LDL: both carry cholesterol through the blood. But Lp(a) has one important difference: it also carries an extra protein called apolipoprotein(a). That added piece makes it more harmful and harder for the body to clear. Unlike LDL, lipoprotein(a) levels are mostly influenced by genetics: around one in five adults have high lipoprotein(a) levels, but most don’t know it. A single gene, LPA, which produces apolipoprotein, confers most of the differences in risk between people.

Researchers are now developing ‘small interfering RNAs’, or siRNAs, to silence the LPA gene specifically. Unlike CRISPR and other gene-editing tools, which make permanent changes to DNA, siRNAs work by targeting the temporary messenger RNA (mRNA) instructions that cells use to make proteins. In many cases, siRNA effects are short-lived; in this case, the effect is unusually long-lasting – a single dose of siRNA can suppress the LPA gene’s activity for months. siRNAs are already approved for treating some rare liver diseases, and now several are being developed to target the LPA gene.

In a recent phase 2 trial, one such drug, lepodisiran, was given to 320 people with very high Lp(a). A single injection of the highest dose lowered their Lp(a) levels by 94 percent and kept them low for months – a dramatic reduction for any cholesterol-lowering therapy.

But while the trial showed that lepodisiran safely lowers Lp(a), it’s not clear yet how much that reduces heart attacks or strokes. Large studies find that cardiovascular risk increases in a continuous, log-linear form as Lp(a) levels rise, meaning people with lower levels are much healthier.

If further trials confirm the benefits, gene-silencing drugs could offer a new way to prevent cardiovascular disease.

The first gonorrhea vaccination program

The UK will become the first country in the world to offer a vaccine program against gonorrhea. Gonorrhea is a bacterial infection caused by Neisseria gonorrhoeae, which spreads through sex and can cause pain, discharge, inflammation, and sometimes infertility. Cases have been rising, and for every antibiotic class used against it, resistant strains have eventually followed.

N. gonorrhoeae is a challenging target: it varies its surface proteins, doesn’t generate lasting immunity after infection, and seems adept at dodging our immune system altogether. Fortunately, the MenB vaccine, developed to protect against meningitis B, which is caused by a close genetic relative, offers some ‘cross-protection’ against gonorrhea. Protection in trials so far has been moderate: one dose offered around 26 percent protection, while two doses boosted that to 33–40 percent.

Projections from Imperial College London suggest that if uptake is high, the program could prevent 100,000 infections and save the NHS nearly £8 million over the next decade.

Subscribe now

A wearable device measures how much breast milk babies drink in real time

Right now, the way doctors and parents estimate how much milk a baby drinks during breastfeeding is surprisingly low-tech: they weigh the baby before and after a feed. This is often imprecise, especially for preterm infants. To solve this, researchers have developed a wearable device that uses four small electrodes placed on the skin to measure milk volume continuously and more accurately.

Breast milk has different electrical properties than fat, skin, or muscle, so when milk flows through the milk ducts, it changes how electricity moves through the breast. The device picks up on that shift. While impedance-based monitoring isn’t new, this version is portable, easy to use, and pairs with a smartphone. In a study of 12 mothers over 17 weeks, the device’s readings closely matched bottle measurements. Even without calibration, its software estimated milk volume within about three percent error.

Montana expands its Right-to-Try laws

Montana legislators passed a new law – called Senate Bill 535 – that expands access to experimental drugs that have passed through only Phase I clinical trials, even if those drugs remain under investigation for FDA approval. Montana already has a ‘Right-to-Try’ law from 2015, allowing terminally ill patients to try medications that had passed Phase I trials – similar to the federal Right-to-Try law passed in 2018. In 2023, Montana broadened its law, letting anyone access those therapies.

This latest bill, passed in May 2025, lets clinics in the state get licensed specifically as ‘experimental treatment centers,’ legally selling drugs and treatments right after their Phase I safety trials. They must also earmark two percent of their net profits each year to help residents more easily access those treatments. One downside of Montana’s approach, though, is that around 90 percent of all drugs that enter clinical trials fail, and most of those failures come at the later stages.

Scientists solve the structures of the ‘sweetness’ proteins

Humans have one main protein receptor that detects everything sweet, from glucose to sucrose and aspartame. Scientists have used cryo-EM imaging to solve the 3D structure of this receptor in tandem with various sweeteners. Cryo-EM works by quickly freezing proteins in their natural state, embedding them in a thin layer of ice. Electron beams are then shot at these samples multiple times, casting shadows to generate precise, composite 3D maps of each protein.

In this case, scientists captured the structure of the sweetness receptor while bound to two sweeteners: sucralose and aspartame. Both of these molecules fit neatly into the same ‘pocket’ within one half of the receptor, known as TAS1R2. The protein structures explain why some sweeteners, like aspartame, are far more potent than natural sugars. Aspartame’s chemical structure allows it to form tighter interactions within the TAS1R2 binding pocket; this molecular embrace stabilizes TAS1R2 in its ‘ON’ state much more strongly than sucrose.

Subscribe now

Contact lenses to see infrared light

The cone cells in the human eye are sensitive to wavelengths from 400 to 700 nanometers. Infrared wavelengths are above that range, and thus humans cannot see them.

A group of researchers, writing in Cell, report that they’ve made contact lenses that convert infrared light into visible light, which can then be detected by the human eye. Each contact lens is packed with nanoparticles that absorb infrared photons at 808, 980, and 1,532 nanometers and then emit visible light. The lenses work without any neural interface or batteries, and supposedly cost about $200 to make. The images remain occasionally blurry and the lenses aren’t great at detecting more subtle infrared sources.

The competition behind AlphaFold is at risk of shutting down

Critical Assessment of protein Structure Prediction (CASP) is a competition that began in 1994 and aimed to track progress against the ‘protein folding problem’ – the challenge of predicting the 3D shape of proteins based on their amino acid sequence. CASP’s goal was to create a standard and independent evaluation of computer models, allowing scientists to properly understand which models were performing better and where gaps remained.

It has been a huge success so far: in 2020, at CASP’s 14th competition, AlphaFold achieved a 90 percent success rate at predicting the structure of proteins, far beyond the roughly 50 percent achieved by other models. In 2024, it achieved a nearly 95 percent success rate.

But further objectives remain. Last year’s competition tested models’ ability to solve the structure of RNA molecules, interactions between proteins and molecules, and the shapes of disordered proteins. Unfortunately, funding from the NIH is now running out, as well as emergency support from UC Davis. Will the Trump administration reinstate funding, or will another funder step up?

On building (and rationing) infrastructure

• Getting clean water to people has been a problem for as long as humans have had permanent settlements. Currently 2 billion people are drinking water polluted by faeces, chemicals and other contaminants and many water systems in the rich world are now over a century old and failing.

• A conversation with Stephanie Pollack, former Massachusetts transportation secretary and Biden-era deputy administrator of the Federal Highway Administration, about why US transit megaprojects take forever and cost fortunes. There’s a strong incentive to give incredibly optimistic numbers – the only way to start a conversation about a $12 billion project is to pitch it as a $2 billion project. The scope creeps up as bridges, parks, and other giveaways are added to the plan and long timelines inflate prices as agencies are forced, by regulation, to buy work in small batches.

• Brian Potter has written a list of 50 things he’s learned writing Construction Physics – not his experience writing a blog – things like ‘Construction has gotten much safer over time: not only have death and injury rates in construction fallen, but they’ve fallen much more than in agriculture or manufacturing.’

• Society needs to crack down on pushy fundraisers.

• Congestion pricing in New York is working. Rush-hour travel time has fallen by 48 percent and complaints about car honking have dropped by 70 percent.

• Cycling in London has increased by more than 70 percent since 2017.

• Britain has become shabbier and more disorderly. Richer people in purpose-built commercial developments pay for their own cleaning, security and maintenance – activities that local governments are notionally responsible for.

• A remote village in Arizona still receives its mail by mule.

Lives saved with gene-editing, antivenom, and deregulation

• Together, a team at the University of Pennsylvania and Danaher, a conglomerate that manufactures medical products, healed a baby with personalized gene-editing.

• In Sub-Saharan Africa antivenom prices are about $124 per dose – ruinous for people earning $2 a day. The market has collapsed into a ‘lemon’ equilibrium in which dubious products erode trust, governments hesitate to lend support and so manufactures won’t scale to bring down costs. Venoms themselves are wildly heterogeneous with different species, regions, and ages of snake delivering needing different responses. But, there has been some movement towards creating a universal antivenom. About 90 percent of lethality comes from just three toxins and high-throughput screening methods can filter candidates down to a few dozen molecules and programs like AlphaFold or Rosetta and assess whether the key proteins will stop working in places without cold storage.

• Statins are basically safe – in the UK you have been able to buy Simvastatin on pharmacy shelves since 2004. Why not deregulate low-dose statins to make them regular over-the-counter drugs everywhere?

  Share

Lead turns to gold, gold turns into R&D, and R&D makes cheaper bacon.

• Physicists can turn lead into gold – for a fraction of a second.

• The US Food and Drug Administration has approved pigs, CRIPSRed to be immune to porcine reproductive and respiratory syndrome, for human consumption. Expect your pork from the Pig Improvement Company in late 2026.

• Across craft, sport and software, speedy feedback loops are what allows us to improve. But biology still runs on slow cycles. Through things like directed evolution, we can already see how a process that is meant to happen on the multi-lifetime-scale can be sped up. To shrink the cycle time we can push for more robotics, invest in predictive models (and open up datasets to make them more effective), and generally call for a culture with less ‘masochistic patience’.

• Spaced repetition systems have gotten better.

• The best horse ever lived 50 years ago. No horse since has come close to Secretariat’s records and he was thought to be well-tempered and easy to train. (Was he doping?)

Geoengineering, robots, and quirky, old hardware.

• Enhanced rock weathering is a method for removing carbon dioxide from the atmosphere. It speeds up the natural process by which silicate or carbonate rocks react with carbon dioxide and water to form stable minerals. Austin Vernon estimates that it can be less than $100 per ton removed at large scale, but only with careful rock selection, aggressive grinding, and clever design.

• Drones can now cause lightning strikes.

• Kyle Vogt, co-founder of Twitch and Cruise – a self-driving car company – has started a new company: he’s building robot butlers. He talks about the challenges scaling robots, why he will never sell another company, and why he thinks Tesla had the right strategy for building autonomous vehicles.

• It is possible to manufacture a complete smartphone in the US but it’s pretty low-spec and costs $2,000.

• The Philips Machine is an analogue computer which uses fluidics to model how an economy works. While designed as a teaching tool, it was found to be surprisingly accurate and was used as an economic simulator.

Gotta catch ‘em all

• How Claude learned to play Pokemon and what it teaches us about LLMs

• AI 2027 tries to sketch out what a world with superhuman AI looks like. And here’s a summary and review.

• Tyler Cowen and Avital Balwit (of Anthropic) discuss how they think our lives will change with AGI. They think we will have more time for leisure, our work will require us to be less agentic, that there will be human-only spaces, and that more people will start to rebel against the encroachment of AI into the things we love.

• In Seeing Like a State, James C Scott popularized the idea that governments can only successfully regulate what they can measure. With AI, more things will become measurable and therefore under the scope of state control. This will likely mean better coordination of things we want, like traffic safety, but will – through things like identify verification, tracking, and administration – also make draconian, anti-liberal measures more likely too.

Subscribe now

Planes in my back yard

• There’s a suburb in California designed so people can conveniently park their personal planes.

  

• Britain Remade has ideas about worker-friendly planning reforms. They recommend granting automatic planning permission in cases where the majority of residents vote to densify their council estates, building new towns near cities with robust job markets, designating regions of low environmental quality to fast-track onshore wind and solar projects, and aligning nuclear regulation with international standards.

• The US has three different kinds of housing problems. First, places where there is too little housing of any kind. Second, places where housing is cheap by national standards but incomes are so low that people still can’t afford them. And third, places where the typical family can afford a home but they have little choice about what type.

• Homes used to be much more uncomfortable. ‘In the Middle Ages, people didn’t so much live in their houses as camp in them’. But incrementally, by introducing more rooms – for privacy, electricity for light, warmth and easy cleanliness, they became pleasant.

Why Chinese solar panels are cheap and British nuclear is expensive

• Chinese batteries and solar panels are cheap because of low labor costs, high levels of automation in manufacturing, and cheap energy.

• Nuclear energy in the UK is expensive because of regulation.

• A French nuclear-fuel giant (Orano) and a Washington DC startup (Curio LV) just struck a deal that could turn spent nuclear fuel into yet more energy – something the US abandoned almost half a century ago over bomb-making fears. Nuclear fuel is also the hip new workout accessory. There are even nuclear raves in Washington DC.

• The Frontier Observatory for Research in Geothermal Energy, or FORGE is a lab conducting research into enhanced geothermal systems. While enhanced geothermal systems are yet to achieve broad commercial viability, FORGE has successfully attracted private capital to geothermal by derisking some of the initial research.

• And if these links aren’t enough here’s a list of sources if you’re interested in keeping up-to-date on energy.

Ideas are easy to find; growth, jobs and a European Single Market, on the other hand …

• This working paper makes the case that economic growth is getting harder to find – not ideas. The authors find that researchers’ patenting productivity is increasing.

• The Single Market, as it is broadly understood, is a myth: mutual recognition of standards fails in practice, the EU directives do not harmonize EU legislation as member states often gold-plate the regulations, and the EU Commission is not enforcing the Single Market.

• New grads are now more likely to be unemployed than the average American. The graph shows the decline started in the early 2010s, so it can’t all be explained by AI.

Parenting across cultures

• Matrilineal descent is rare in human societies that keep cattle. This paper shows that the switch to patrilineal descent follows cattle, not the other way round.

• How do parents maintain influence over their children when they move across the world? They try to teach them religion and choose their spouses for them, but this has mixed results.

• There are still stark differences in health between eldest sons and other siblings in Indian families. And, as families in India shrink, if a preference for having at least one son persists, then we might see an even greater desire for sex selection.

What a way to make a living

• In defence of email jobs

• This man increased his productivity by hiring someone to sit with him while he works.

• The world’s first org chart

Printing books, silica gel, clothing, Australian state capacity, and more progress stuff

• See inside the factory printing hardback copies of Pride and Prejudice.

• Why are silica gel packets everywhere? It’s because we now ship and store things that are vulnerable to dampness. Silica gel is a sign of globalization.

• US households now spend only 3 percent of their budgets on clothing, down from 10 percent in the 1960s — we own more clothes but they are so much cheaper. The general consensus that during this time the quality has fallen. Making thick and durable clothing that a person only wears for a season would be a waste of resources. In cases where performance matters, like running shoes, PPE, and insulated hiking clothes, quality has soared. High-end classics are still durable, they’re just expensive.

• Australia has unusually good state capacity. This literature review says it’s because Australians are obedient to impersonal authority making it easier to enforce policies like compulsory seat belts, because they trust their institutions which means they don’t suffer from American-style adversarial legalism or public service risk-aversion, because Australia has high quality, respected organisations that produce independent policy research, and because it pays its public servants well.

• A list of all the organizations that are part of the progress movement.

• If you live in Dublin, consider joining The Fitzwilliam Reading Group.

In today’s edition of Links in Progress, Coby Lefkowitz returns to review more beautiful developments built around the world in the last two decades.

This issue of Links in Progress is dedicated to practicing and aspiring architects, developers, planners, policymakers, public officials, and the general populace, who are interested in creating a better built environment in their own communities. Enjoy!

LC Line & Low

• Location: Charleston, South Carolina

• Year Completed: 2025 (Completion later this year)

• Building Type: Mixed-Use

• Unit Breakdown: 293 apartments, 22,000 square feet of commercial space

• Architect RAMSA

• Developer: Lifestyle Communities

Most mid-rise apartment buildings have come to be looked at with derision in recent years among the general public, due to the cheapness of their construction, their anti-urban orientation, and dreary (while also incomprehensible) designs. This new development in Charleston transcends all of what most people are trained to think of when they visualize a mid-rise. Robust, expertly detailed, and coherent. It is a gold standard for what new development can be.

Subscribe now

Fouquet's New York

• Location: New York City

• Year Completed: 2022

• Building Type: Hotel

• Unit Breakdown: 96 key-hotel, 2 restaurants, 2 bars.

• Architect SBJ Group

• Developer: Caspi Development

Blending in seamlessly with the historic character of New York’s Tribeca neighborhood, Fouquet’s is a gorgeous composition of brick, stone, and cast iron elements. As a luxury hotel, one would expect a high quality suite of finishes, but somehow the work of Stephen B Jacobs Group surpasses reasonable expectations for what a new structure should offer.

El Andaluz

• Location: Santa Barbara, California

• Year Completed: 2009

• Building Type: Mixed-Use

• Unit Breakdown: 7 residential condos, 2 retail units

• Architect: Jeff Shelton

• Developer: n/a

Jeff Shelton is the best practicing architect in America today. What do I actually mean here? Simply that Shelton consistently creates structures that are captivating and whimsical, and top in class on the marks of liv-and-lov-ability. Beyond this, they are so full of spirit they have the opportunity to do what precious few structures in North America can – endure with affection. El Andaluz is one of his studio’s largest projects (if not the largest), but unlike many other firms, his work improves as it scales. His signature details have not been diluted with size. If anything, extra attention is given to them to ensure size does not obviate,

Share

Palma Social Housing

• Location: Palma, Mallorca

• Year Completed: 2022

• Building Type: Multifamily

• Unit Breakdown: 8 apartments

• Architect Miguel Nevado, Alfonso Reina, Antonio Martin, Carles Oliver, Xim Moya

• Developer: Balearic Housing Institute

It is rare to see high quality, robust construction in new housing developments today, much less public housing which has become nearly synonymous with cheap and thoughtless composition. It’s in this context that this warm, earthy, and eminently welcoming social housing in Palma is made all the more remarkable. Marès sandstone courses through these 8-units, cladding enchanting barrel vaults and walls, accented by lovely wood.

The Red Roof

• Location: Quang Ngai, Vietnam

• Year Completed: 2019

• Building Type: Single Family Home

• Architect: TAA Design

• Developer: Private Client

A home in a smaller city along the central coast of Vietnam, The Red Roof punches above its weight relative to its size and location. At just over 800 square feet, this home’s impact comes from its two most defining features: a stepped roof garden, and red plaster facade, which warm and enliven the street. As the small lot denied the possibility of creating a large yard space, the architects opted to move the gardens to the roof, and enclose other outdoor spaces in a series of small courtyards. This creates a level of intimacy that’s rare in single family homes, without crowding uses together.

Mass Timber Parisian Courtyard

• Location: Paris

• Year Completed: 2020

• Building Type: Multi-family building

• Unit Breakdown: 14 apartments

• Architect Mars Architectes

• Developer: Gecina

Located on the site of a former parking lot in the courtyard of a Parisian apartment building in the 12th arrondissement, Mars Architectes have delivered one of the more attractive, and welcoming, buildings of recent years. The warmth of the wood and the consideration of the landscaping combine to create a harmonious and supremely compelling product. It is a reminder that every patch of land has the ability to completely transform our built environment

Grace Farms

• Location: New Canaan, Connecticut

• Year Completed: 2015

• Building Type: Public / Cultural Venue

• Architect SANAA

• Developer: Grace Farms Foundation

Carving through the hillsides of Southwestern Connecticut, Grace Farms is one of the more surprising structures one is likely to come across. The acres of large, well manicured plots of this staid suburb don’t at first seem to make the best neighbors for a bold architectural foray. Maybe the contrast is what makes SANAA’s building so appealing. Adorned with glass, steel, concrete, and blonde woods, it is a work of grace, floating above the 80-acre public nature reserve. I have had the fortune of visiting Grace Farms several times, and fear I can’t give written justice to the tranquility that washes over one whilst spending time there.

Share

Little Island

• Location: New York City

• Year Completed: 2021

• Building Type: Park

• Architect: Heatherwick Studio

• Developer: Hudson River Park Trust and Pier 55 Project Fund

Structures that don’t have a functional commercial or academic use have become the subjects of derision among a subset of critics of architecture and political economy. But for the millions who visit this park Little Island is a triumph. Expanding the usable square footage of New York’s park system by extending into the Hudson River on 132 bulbous concrete columns, it remains a surprising delight every time one passes by. Without parks, we are lesser beings. New Yorkers (myself included) are better for the existence of Little Island. May a thousand further islands (preferably with housing!) bloom.

Dapi Mountain Restaurant

• Location: Xun County, Henan Province

• Year Completed: 2023

• Building Type: Restaurant

• Architect Galaxy Arch

• Developer: n/a

With its sweeping roofline, natural materials (like local grey stone from Taihang Mountain, the same wood used in local historic building, and slate roof tiles), and seamless integration into a historical style, Dapi Mountain Restaurant is an escape into beauty and harmony. The structure contours itself, and thus blends, into the natural environment, seeming as though this place on the mountain top was waiting especially for it. If only all structures availed themselves so well to bounty of their local contexts.

Coby Lefkowitz is a real estate developer and writer based in New York.

1. Biologists have long used ‘reporter’ genes, such as green fluorescent protein, to watch molecules move around inside living cells. But fluorescent proteins can only be seen from close up. MIT scientists have solved this problem by creating ‘hyperspectral reporters’, which are DNA sequences that instruct bacteria to make pigment molecules that subtly shift light in the hyperspectral wavelengths, thus creating unique fingerprints. When these hyperspectral reporters are inserted into soil bacteria, the bacteria can be seen using cameras mounted to drones from up to 90 meters away. Researchers can program cells to sense explosives, trigger reporter genes, and then spot these areas from afar.

2. Screwworms have returned to Central America. Costa Rica confirmed its first screwworm infection in over 30 years on February 7 2024, when physicians found dozens of larvae in an 80-year-old Canadian visitor’s leg ulcer; Toronto surgeons removed more than 40 larvae in two procedures. And earlier this year, physicians reported at least 30 cases of screwworm infections in Nicaragua. For over 50 years, the USDA has airdropped millions of sterilized male screwworm flies weekly along the Panama–Colombia border. Female screwworms only mate once in their lifetime; these sterile males mate with the females, but produce no offspring, thus slowly reducing insect populations. This ‘sterile-insect technique’ eradicated screwworms in the U.S. in the 1960s, when the pests killed cattle and inflicted about $100 million in annual losses. But now, the ‘screwworm wall’ between North and South America has been broken.

3. Scientists have developed a new gene-editing tool, STITCHR, to insert large pieces of DNA into genomes. Existing tools like CRISPR–Cas9 can make small edits easily but struggle to insert big pieces of DNA efficiently, which has limited efforts to replace entire genes or fix complex mutations. To find a better tool, researchers searched across thousands of species and identified R2Tg, a highly active retrotransposon from the zebra finch. They engineered it into a programmable system that can install edits up to 12.7 kilobases long – roughly the size of the average human gene – without leaving mistakes behind.

4. Scientists have developed a portable test to detect a tuberculosis infection from blood or respiratory samples. Tuberculosis (TB) is the largest infectious disease killer worldwide, killing around 1.2 million people annually, mostly in poor countries, where fast, accurate diagnosis is often hard to get. Researchers have built a handheld device paired with a ready-to-use tube that processes the sample, looks for DNA of Mycobacterium tuberculosis, and gives a simple readout. In tests on children and adults, the system detected a higher share of TB cases than standard methods (meaning it has ‘high sensitivity’). It also correctly ruled out TB in most people who didn’t have it (‘high specificity’), at levels that met World Health Organization targets.

Subscribe now

5. Researchers have created the largest ‘brain wiring diagram’ to date. Researchers first mapped all 302 neurons in the C. elegans worm back in 1986 to link physical wiring with behavior. Now, a consortium of scientists has traced the largest connectome yet – about 75,000 neurons and 500 million synapses in a mouse, out of roughly 7 million total neurons.

6. DNA vaccines are easy to produce, stable at higher temperatures, and faster to design than many other types of vaccines. But they’ve usually needed special devices to push the DNA into cells, like ‘gene guns’ or electrical pulses. An alternative option was to include DNA vaccines within lipid nanoparticles (LNPs) – the same technology used to deliver mRNA COVID-19 vaccines – but until now, this hadn’t worked well for DNA vaccines. Scientists have created a new DNA-LNP formulation that better packages and protects DNA, helping it enter cells and activate the immune system. With this method, they developed new vaccines for influenza and COVID-19 and tested them in mice and rabbits, which showed strong antibody and T cell responses that lasted for over a year in some cases. This approach could be widely used to create simpler, more stable vaccines for various diseases.

7. Drug companies are pooling data to improve AlphaFold3. AlphaFold 3 is Google DeepMind’s latest AI model for predicting protein shapes from genetic sequences. It was trained on over 200,000 protein structures, but it has few examples of drug–protein complexes. To fill this gap, pharmaceutical companies – including AbbVie, J&J, and Sanofi – are pooling their proprietary protein–ligand structures into a secret dataset off-limits to academics. They’ll retrain an open-source AlphaFold 3 replica to predict both protein folds and how candidate drugs bind. Hopefully this proprietary model will be able to virtually optimize drug candidates while requiring far fewer experiments in the laboratory.

8. For decades, the FDA required drug developers to prove safety in animals –often dogs or non-human primates – before first-in-human trials, despite animal data frequently failing to predict human outcomes. But the agency recently unveiled a detailed roadmap to phase out, or dramatically reduce, animal testing, allowing companies to submit AI or organoid data instead. At first, this roadmap applies to a class of drugs called monoclonal antibodies, but only for those used to target non-human antigens (like bacteria). Monoclonal antibodies are some of the most common drugs on the market and include bestsellers like adalimumab (Humira), pembrolizumab (Keytruda), and dupilumab (Dupixent). The FDA is also offering case-by-case waivers for other types of drugs.

9. The FDA approved a new long-acting hemophilia treatment. Hemophilia is a rare genetic condition that slows down blood clotting, which can lead to frequent bleeding. For years, patients have relied on regular infusions of clotting factors or similar products to prevent or treat bleeds. These can be inconvenient and time-consuming, especially for people who need frequent dosing. In March, the FDA approved a new treatment called Qfitlia, a shot under the skin that’s taken as infrequently as once every two months, and in trials, it reduced bleeding episodes by over 70% compared to standard on-demand treatment.

Share

10. A new lab-based study found that nitisinone, an existing FDA-approved drug, can kill malaria-spreading mosquitoes after they feed on treated blood. Malaria is spread by Anopheles mosquitoes, which need to digest large amounts of protein after feeding on blood. As part of this, they break down the amino acid tyrosine (one of the building blocks of proteins), which is toxic if it builds up. Nitisinone, a drug used to treat rare metabolic disorders, blocks the enzyme that mosquitoes rely on to process tyrosine. When female mosquitoes fed on blood samples from patients taking this drug, they died, even if they were resistant to standard insecticides. This approach is similar to how ivermectin, a drug used to treat parasitic infections, can kill mosquitoes that feed on people taking this drug, and has also been used in anti-mosquito bed nets.

11. Researchers developed a potential Lyme vaccine that prevented the disease in mice. Nearly 500,000 Americans contract Lyme disease each year from tick-transmitted Borrelia burgdorferi, and the only FDA-approved vaccine was withdrawn in 2002 after poor sales due, in part, to vaccine skepticism, and its inconvenient, three-dose schedule.

12. Researchers are also making headway against another major bacterial threat: Clostridioides difficile, a bacterial infection that often strikes after antibiotic use, causing gut inflammation and serious diarrhea, especially in hospitals. Past vaccines targeting the bacteria’s two major toxins, TcdA and TcdB, have failed. Now, researchers have developed a mRNA vaccine encoding parts of the bacteria’s toxins and a protein that helps the bacteria move. Injected into mice in two doses, the vaccine triggered strong antibody responses: all unvaccinated mice died, all the vaccinated mice survived.

13. In the United States, about 3 million antimicrobial-resistant infections occur each year. One way to counteract this problem is to discover new classes of antibiotics that microbes have not yet evolved resistance against. Researchers recently identified a new antibiotic with a novel molecular mechanism. Lariocidin, a small, knotted protein naturally produced by Paenibacillus bacteria, is effective against both Gram‑positive and Gram‑negative bacteria – including multidrug‑resistant Acinetobacter baumannii – and, in laboratory experiments, researchers did not detect any toxicity to human cells. It will need to go through further clinical testing and, like other new antibiotics, will likely face major economic hurdles in reaching patients, as limited financial incentives have slowed antibiotic development for decades.

14. What may be the first cure for a prion disease was recently given permission by the FDA to start human trials. Eric and Sonia Minikel have been researching prion disease for around a decade, hoping to find an effective treatment for Sonia Minikel’s inherited prion disease, fatal familial insomnia (FFI). With improvements of gene-editing technology, they developed a potential drug that silences the prion protein in infected mice.

In defence of cars, sprawl and natural disasters

• With careful planning, roads and cars can be an efficient means of transport even in dense cities. For example, ring roads are often inefficient for drivers because, instead of going directly from A to B, they need to go round the urban centre they are trying to reach. But they spare city centres from traffic so they can be efficient in terms of land use. To make ring roads fast enough that they are appealing to drivers in London, we would need 44 lanes (ideally spread over multiple ring roads, not as one 44 lane monstrosity).

• Why America should sprawl (as well as build up).

• There used to be a silver lining to natural disasters – they cleared the way for new development. But today’s rules mean that rebuilding is often more constrained than original construction.

• Tract was a venture-backed company built to address Britain’s housing crisis by improving the planning permission process. Its founders have published a post-mortem on why they are closing the business and what went wrong: the conservatism of the property market, the lack of scale in Britain, and their unforced errors.

• Texas has very cheap housing by American standards, in part, due to their special Municipal Utility Districts which allow developers to build infrastructure upfront and have future residents repay the costs over time.

• National elections are rarely swayed by local infrastructure, so it is rarely a priority for national politicians. However, local elections can be decided by local conditions so devolving infrastructure decisions to mayors could mean more and better infrastructure in the UK.

• You can watch London’s trains move around in real time on an interactive 3D map.

How hypertumors stop whales from getting cancer, and other science

• As animals get bigger and have longer life spans they have more cells which go through more cell divisions. We should expect this to mean the cell’s DNA in larger animals has more copying errors and they get more cancer. Instead we see the opposite. There are two main theories why. First, cancerous tissues are themselves more vulnerable to getting cancer. A creature like a blue whale can survive having gram-sized tumors that grow rapidly and die. In a smaller animal, a similarly sized growth would kill them. Second, we have found evidence in elephants of genes that are protective against cancer itself. It is plausible that similar adaptations exist in all large animals.

• You may have seen the news that Colossal Biosciences has genetically engineered the first dire wolf to live in over 10,000 years. This isn’t strictly true, but the science to make these new dire wolf-like pups is impressive nonetheless.

• Heather Stone is a science policy analyst at the Food and Drug Administration who want to improve drug access for rare and deadly diseases. She built CURE ID, a platform for reporting and sharing case studies but this has been largely ignored by the administration’s bureaucracy. But Stone ignored her own CURE ID system and personally fast tracked approvals for the drugs a five-year-old girl needed when she was infected with Balumuthis. This piece is part of a Washington Post series on how different government officials work.

• Comparing the lifespans and rates of aging in different animals is difficult. For example, we typically count a zebrafish’s age from when its zygote is first fertilized but we count tree shrew ages from when their eyes first open (21 days after birth).

• We published a piece on extending fertility windows to help women ‘have it all’. Demographer Lyman Stone wrote a critique of what he calls ‘techno-optimism about fertility’ (paywall). Original author Ruxandra Teslo responded with a critique of Lyman’s critique and Lyman duly responded again. May the debate continue.

• Edwin Chon developed methods for plasma fractionation – extracting useful proteins like albumin from blood – and scaled it up to treat wounded soldiers in World War Two. Under the pressure of the war he built a team of chemists, engineers and technicians that iterated quickly to build something that could work on the front lines. After the war he returned to fundamental research and deeply regretted it because without the urgency of wartime, his lab lost much of its dynamism.

• By living healthier we minimize the economic problems that come with low birth rates and aging societies. Silver Linings models what impact different anti-aging therapies will have on the economy.

Where’s my flying Sherpa?

• No one is safe from automation (or productivity-enhancing cutting-edge technology). A Nepali startup is making drones to help Sherpas.

• It is easy to believe that AI will change the world dramatically but it is difficult to have a vivid sense of what this future could be like. A history of the future (2025-2027, 2027-2030, and 2030-2040) paints a picture of what the world may look like – humans in the loop managing AIs that they understand less and less each day, entire countries going AI free (except for defence) and some guesses at what this implies.

• Wikipedia is complaining about the cost of being repeatedly scraped by internet crawlers. It says ‘our content is free, our infrastructure is not’. Similarly, Google’s AI summaries are decreasing traffic, and therefore revenue, to people who collate and share high-quality factual content. Google is trying to work out if it can afford to keep upsetting writers and, as an experiment, it removed European news results for one percent of users in eight European countries and found it had no measurable impact on revenue.

• 23andMe has filed for bankruptcy, its CEO has resigned, and it is facing fines for data breaches.

• The UK’s national AI institute is a failure.

• Why hasn’t AI for pathology taken off?

Stonehenge upgraded

• Pictures of a concrete-based remodel of Stonehenge, which really exists outside Maryhill, WA, that fixes the ‘problems’ with it.

• Scott Alexander writes about slave morality vs master morality through the lens of different online commentators. Scott says that ugly postwar buildings are part of our post-WW2 ‘slave morality’ or ‘ensmallening’.

• History of Industrial Design is a lecture series that explains the relationship between art and industry. Watch if you are curious about where chairs came from.

On building a working state, marriage and babies, British liberalism and, of course, tariffs

• There is a common misconception that the United States stopped building nuclear reactors after the 1970s. But that is not the full story: the US Navy has consistently built at least one small nuclear reactor every year since the 1950s. Read about how Admiral Hyman Rickover created the surprisingly functional Naval Reactors Group.

• Europe has expensive energy and Ireland’s is the most expensive. Here’s one reason why.

• The Mercatus Centre says that most tax breaks distort the economy, create unfair advantages, and make the system more complex. A few serve a legitimate purpose, like avoiding double taxation, but most don’t.

• The Centre for British Progress is a new think tank, based in London. Their founding essay explains that they believe material progress drives human progress but that today wages, productivity and investment are all flatlining. They want the UK to have a more entrepreneurial state that embraces things like Small Modular Reactors and redesigns institutions to reward initiative and streamlines regulation.

• Divorce rates for most rich countries peaked whenever their divorce laws were first liberalized but the US remains an outlier with persistently high levels.

• The End of Children shows what a world with collapsed fertility looks like: schools with three pupils, day-care centres turned nursing homes, and a young population who see the culture as one in which it is unimaginable to raise children.

• The Liberal Digest is a new Substack promoting the best of British liberal thought.

• America has everything-bagel liberalism. Britain has everythingism: the belief that every proposal, project or policy is a means for promoting every national objective, all at the same time.

• Confused by the economics of tariffs? This post may help.

Books: What virtues should your employees, scientists and politicians possess?

• I read Tyler Cowen’s book on Talent. It is focused on hiring but mostly prompts you to think about what traits to value in other people and how to work out if someone possesses them. I recommend it to anyone hiring or dating – so it was basically useless to me but I hope it finds a broad audience.

• In 2000, Jan Hendrik Schön fabricated data to pretend he had created a powerful new method for turning organic crystals into transistors. After being exposed, the scientific community – including Nobel prize-winners and major journals – continued to support him.

• Published in 1985, Amusing Ourselves to Death predicts a world in which politics is taken over by enraging and entertaining reality TV stars.

Earlier this year, the UK Government announced its intention to create a Nuclear Regulatory Taskforce to pull down the barriers holding back the UK nuclear industry. It was reported this week that John Fingleton, former chief executive of the Office of Fair Trading, is set to be named as the taskforce’s chair.

As the taskforce gets to work, we present a reading list we’d recommend for its members, or anyone looking to better understand the importance of nuclear and how to get back to cheap, safe and fast nuclear power, wherever they are in the world.

1. In Georgia, pressurized water reactors Vogtle 3 and 4 came online in 2023, arriving seven years late and $17 billion over budget. This apparently was driven by a shortage of experienced personnel, incomplete design, a new regulatory certification process, supply chain delays, and the bankruptcy of the original engineering, procurement, and construction provider. This is nicely documented by this deep dive. The Clean Air Taskforce also has a punchy set of Vogtle lessons. For a broader look at the problem of why nuclear power plants cost so much to build, Brian Potter at the Institute for Progress has you covered.

2. For an account of why the nuclear industry failed to live up to its 1950s promise of ‘energy too cheap to meter’, read Jack Devanney’s Why Nuclear Power has been a Flop. Devanney tells the story of how weak scientific evidence and a regulatory ratchet drove up the cost of building new nuclear power plants. The chief villains of the piece are the Linear No Threshold hypothesis (LNT) and a regulatory standard known as ALARA, or As Low As Reasonably Achievable. These are the interlinked notions that there is no ‘safe’ level of radiation exposure (whether it’s from background radiation, normally operating power plants, or releases of radioactive materials during accidents), and that harm scales linearly with dose, therefore any ‘reasonable’ measure that can reduce exposure should always be implemented. Devanney makes a compelling case that all three elements of that statement aren’t true. You don’t have to agree with all of Devanney’s analysis (and I don’t) to get significant value from it. Jason Crawford has a helpful summary over at Roots of Progress.

Share

3. While the US has been the world’s biggest net export of bad regulatory principles around nuclear, there is a distinct UK element to the story. This seems to have come from a prioritization of flawed indigenous designs over better international ones, bad project management, and a botched privatization which cost the UK its former nuclear leadership position. I’m less down on the advanced gas reactors than the author, but think this is directionally correct.

4. In the long-term, we need to make it easier to build new nuclear power stations, but extending reactor life and reviving old power stations offers a short-term fix. Both of these pieces drive home the relatively low costs involved, the short timelines to bring plants back online, and the significant economic upside. For example, there are nine reactors in Germany that are still in the early stages of decommissioning – reviving just these would meet 18 per cent of the country’s energy needs.

5. Maybe smaller reactors are beautiful, although we’d need a series of regulatory changes we’d need to smooth the path to their deployment. Many of the measures proposed in this paper from Eamonn Ives, including mutual recognition of designs with trusted regulators, streamlining environmental permitting, and cutting consent timelines also align with recommendations in the Tony Blair Institute’s report on unlocking nuclear power for AI data centers.

6. Either way, smarter regulation is part of the solution. Part of this will come from general improvements to the cumbersome system of environmental consents that gum up planning and streamlining regulatory approvals. There’s also scope for more creative reforms as well. For example, could the US consider transferring some elements of nuclear regulation, which don’t concern containment and proliferation, to the states to facilitate healthy competition?

7. Much of the regulation that determines safety standards on radiation stems from decades-old controversies about fall-out from nuclear weapons testing. Elements of Controversy provides a scrupulously fair history of the Atomic Energy Commission. The AEC was the first US nuclear regulator and had the awkward dual mandate of promoting and regulating both nuclear power and weapons testing. Its bias towards secrecy helped fuel anti-nuclear sentiment among many scientists, even though there’s little evidence that fall-out from weapons tests caused real harm to the American public. It’s a fascinating case study in how (not) to navigate on emotive scientific questions.

8. If you need a break from the intricacies of nuclear power, it’s worth remembering why cheap and abundant energy matters. In short, energy is the fundamental constraint on humanity's physical capabilities, determining what we can build, transform, and accomplish in the material world. Abundant, cheap energy would enable us to overcome environmental challenges, eliminate resource scarcity, and unlock technologies that seem magical by today's standards, according to Benjamin Reinhardt in a special edition of Works in Progress.

Alex is an editor at Works in Progress, focused on AI and energy. He’s also the author of Chalmermagne, a Substack covering technology, policy, and finance.

1. Twenty US states passed 65 housing supply bills in the year to June 2024. Colorado’s transit-oriented development (aka ‘TOD’) pre-emption – requiring zoning for 40 dwellings per acre on average within a half mile of frequent transit – is arguably the most significant TOD reform passed in the US to date, at least pending details on enforcement.

2. NYC housing plan enables 82,000 homes over the next 15 years beyond the existing pace of development. The final version will create the largest 'parking freedom' zone in the US, eclipsing the citywide parking reforms in the geographically small cities of Austin, San Francisco, and San Jose in total acreage. Next steps include neighborhood upzonings and the Manhattan Plan.

3. 'HUD Code' homes, following rules written by the Department of Housing and Urban Development to regulate factory-built homes, cost just $90,000 to build. But until this year, the HUD code was restricted to single family units, limiting their market appeal to areas with low land costs. Last year’s new rulemaking and upcoming legislative reforms together are expected to change this.

4. Up for Growth’s regular annual tally of regional US housing underproduction asks how many homes would be needed in the short run to alleviate overcrowding and achieve a healthy vacancy rate assuming no movement between cities. By this measure the US is missing 3.9 million homes. But ‘housing need’ estimates face a basic knowledge problem and this number is a huge underestimate.

Papers of Note:

5. A recent Terner Center report looks at how six US states use funding access and priority grants to encourage municipalities to upzone. Massachusetts, California, New York, and New Hampshire use voluntary programs. Montana and Utah enforce local participation. They target zoning reforms, ADUs, and affordable housing. Unfortunately the states have not evaluated its impact on actual housing production.

6. The desire to see property values rise is a common explanation for NIMBYism. But we often see restrictions on housing supply in areas where growth would overwhelmingly raise land values.

7. Traditional office to residential conversion requires expensive whole-building renovation. Large post-war office buildings with deeper floor plates generate either windowless bedrooms or require the excavation of light courts to bring windows into the building’s core. Co-living can cut office to residential conversion costs by 25 percent to 35 percent versus traditional conversion. In Denver, converted co-living units would cost just $123,000 per unit compared to a typical subsidized low-income housing studio costing $400,000 each.

8. In 1986, the US imposed a harsh depreciation schedule on corporate structures. Knowing it would reduce housebuilding, Congress created the Low Income Housing Tax Credit to subsidize non-market housing. With the Tax Cuts and Jobs Act coming up for renewal in this year, the Tax Foundation argues we should extend this pro-growth 'full expensing' logic to all buildings.

9. Condo construction in the US has collapsed in recent decades. Intended to protect owners from poor development practices, strict condo defect laws penalized builders who construct them. Decade-long lookback periods and incentives to sue are common, and mortgage access for condos is tight compared to financing for single family homes. (Meanwhile, Canadian YIMBYs tell me new multifamily housing in Toronto and Vancouver is overwhelmingly condo, and locals complain about the absence of purpose-built rentals.)

1. Nanopore sequencing is a promising, low-cost alternative to traditional genome sequencing methods. Nanopores are tiny holes in a membrane. When a molecule passes through a nanopore, it disrupts an electric field. Fluctuations in the field can be ‘decoded’ to figure out whether a passing molecule was an adenine, thymine, or something else. Unfortunately, its use has been limited by poor spatial resolution. Roche has developed a new sequencing method called Sequencing by Expansion (SBX), which expands DNA molecules into larger polymer structures called Xpandomers and makes them easier to read with nanopores. If successful, SBX would improve the accuracy of nanopore sequencing and bring costs down.

2. Justin Sandefur and Charles Kenny estimate that US foreign aid prevents around 3 million deaths per year at a cost of only 0.2 percent of the US’s GNI. The Trump administration has scrapped 90 percent of all USAID contracts, including some of the most impactful programs. The scrapped contracts seem to include all tuberculosis programs, programs supplying essential medicines for maternal and child health; programs providing HIV treatment; UNICEF’s polio vaccination program; contracts for disease outbreak prevention, detection, and response; emergency food aid; earthquake assistance and more.

3. The FDA approved a non-opioid painkiller. Journavx, developed by Vertex Pharmaceuticals, targets a protein called the NaV1.8 voltage-gated sodium channel, which transmits pain signals from sensory neurons to the brain. No other drugs on the market target this protein; existing painkillers instead bind to several other different NaV channels at once. In a phase 3 trial with 2,000 patients undergoing surgery, Journavx reduced pain as effectively as hydrocodone plus paracetamol and had fewer side effects. It also does not appear to be addictive. But the drug will cost about $233 per week; opioids cost about $12 per week.

4. The FDA approved a new vaccine for chikungunya, a mosquito-borne disease that can result in joint pain for months. The new vaccine was developed by Bavarian Nordic, the same company that manufactured the Jynneos Mpox vaccine. Bavarian Nordic have been rewarded for the results of their chikungunya vaccine with a ‘priority disease voucher,’ which will allow them the opportunity to jump the FDA review queue with another drug. Because they primarily affect poorer countries, drugs for curing tropical diseases face a limited financial market. The ‘priority disease voucher’ acts as a reward for Bavarian Nordic, who can use it to get their next drug to market faster.

5. A guide to ‘forensic metascience’. James Heathers, a researcher and scientific sleuth, has published a textbook explaining how to probe academic studies for potential errors, questionable research practices, and scientific fraud.

6. Chinese researchers made mice with two genetic fathers. They took sperm from male mice and reprogrammed them into stem cells, effectively ‘rewinding’ their development. Those stem cells retained only one set of chromosomes, like normal sperm, but were then fused with a second sperm’s DNA — resulting in both XX and XY sex chromosomes. The resulting DNA was injected into an egg without a nucleus, producing embryos with two paternal genomes. Out of 164 attempts, seven mouse pups were born; they survived to adulthood but had serious health problems and were infertile.

7. Arc Institute launched the largest AI model yet for biology. Called Evo 2, it was trained on 9.3 trillion nucleotides of DNA from 128,000 organisms. Evo 2 can predict whether a given mutation in a DNA sequence will be pathogenic with over 90 percent accuracy, even in instances where scientists have not previously studied that mutation in the laboratory. The model was trained on 2,000 NVIDIA H100 GPUs – about 150 times more compute than AlphaFold. Asimov Press covered the story.

8. In 2022, the retraction rate across hundreds of journals slightly exceeded 0.2 percent, but about 80 percent of these retractions came from journals owned by a single publisher: Hindawi. This latest Nature report shows that the problem is pervasive for specific publishers and for select universities. Jining First People’s Hospital had a five percent retraction rate between 2014 and 2024, which is 50-times the global average. In 2023, 60 percent of retractions were from Chinese institutions.

9. Enzymes speed up chemical reactions. They often do so by going through multiple steps to complete a reaction cycle. However, designing new multi-step enzymes has been challenging because AI tools typically predict static protein structures rather than capturing their dynamic movements. To overcome this, scientists combined AI models RFDiffusion and AlphaFold2, as well as a custom tool called PLACER, to create a “multi-step enzyme” for the first time. Though these AI-designed enzymes are much slower than natural ones (10^-3–10^-1 vs. 10²–10⁵ reactions per second), this proof of concept opens the door to engineering entirely new enzymes that do not exist in nature.

Science

Are seed oils actually bad for you?

Edward Cohn was a protein chemist who discovered and quickly scaled a new way of providing soldiers with blood-derived treatments during World War Two.

A man survives 100 days with an artificial heart. (Not as long as the 10 years a person can survive with a donor heart, but longer than anyone has survived with a pig heart.)

Evolutionary psychology is often criticized for telling ‘just-so’ stories. Adam Hunt and Adrian Jaeggi published a new paper outlining their framework for analysing evolutionary hypotheses, using autism as a case study. If academic papers aren’t your thing you can read Adam’s blog post.

Metascience

The Tony Blair Institute has published a report on how to accelerate scientific discovery in the UK in an age of AI .

In defence of weird research.

People expect universities to serve lots of roles in society – scientific research, training and credentialing people for the work force, inventing technology, creating informed citizens. Ben Reinhart argues that universities would be better if we unbundled them.

Tech

Are better AI models actually better?

Faculty AI have made a new font based on old street signs in London.

Why don’t you have a robot butler?

We have autonomous vehicles in San Francisco and Phoenix but not New York and London? Why is it harder to make robotaxis in cold weather?

How to use AI in your writing.

Reflections on Palantir – its culture, morality, and how it works – from someone who worked there for eight years.

Will MacAskill, one of the founding fathers of Effective Altruism, returns to the 80,000 Hours podcast to discuss what a world with superintelligence is going to look like.

Traditionally biotech innovations came from startups that were then bought by Big Pharma. Now AI-aided drug discovery could shift innovation to different organizations, maybe to Big Pharma which may benefit from better data sets, increased process power, and vertical integration.

China is far ahead of the west on robots – installing more, building them cheaper with a mostly localized supply chain, and with a clear plan to replace human labor.

Housing

The Irish Department of Housing is considering proposals to allow people to build cabins and modular homes in their back gardens without planning permission.

Research shows that four to six story apartments with a single set of fire stairs are no more risky than two stair apartment blocks.

Hundreds of roofs in Gujarat, India have been painted white to keep residents cooler in the hottest months.

‘Affordable housing’ is a misleading moniker. We should call it what it is – subsidized – and deal with the tradeoffs that implies.

Fertility

People usually use total fertility rate as a way of gauging how many children are being born. But it doesn’t tell you everything.

A team are trying to create a new sport: sperm racing. They’re building a microscopic race-track and they’re going to be taking bets. Hopefully the data they gather will help us understand more about male fertility.

Despite having the highest birth rate in Europe, there were more people born in France in 1946 than 2024.

The child penalty is a measure of how much, on average, a woman’s earnings decrease when she has a child. In the rich world, countries with higher child penalties have lower birth rates.

Misc

We live like royalty and we don’t even know it.

How fantasies about future technology can stymy progress in the present.

Japan gets a lot wrong – per capita incomes have barely increased since 1990. But it’s still a nice place to live. What are they getting right?

Matthew Yglesias argues that propaganda is overrated and misinformation mostly confuses your own side

How did footbinding end?

People often say that Europe is less innovative than America because of cultural reasons. Alex Chalmers argues that these ‘cultural differences’ are downstream of other, primarily economic, factors.

An OECD report finds that small energy price shocks can be good for the economy (probably because it hastens the death of unproductive firms) but that severe shocks are not.

Why regulators should have red teams to argue, internally, against bad regulations.

How postcodes work in different countries.

Too cheap to meter is a film that makes the case for abundant energy and explains why nuclear energy – fission and fusion – is our path to get there.

Books

Why Nothing Works: Who Killed Progress―and How to Bring It Back is a new book from Marc J Dunkelman. He talks about the contradiction between the left’s ‘Hamiltonian’ impulse to give the state more power to enact good works and their ‘Jeffersonian’ impulse to protect citizens from whims of people in power. He argues that these contradictory impulses lead to a system of vetocracy in which it is difficult to get anything done like build infrastructure, fight climate change, or have a working welfare system. If a book is too long for you, you can listen to Marc summarize his thesis on Matthew Yglesias’s podcast.

South Korea has the lowest birth rate in the world with a total fertility rate of only 0.7 babies per woman. There are probably a number of reasons why this is the case but one notable reason is poor gender relations – young men in South Korea are increasingly right wing and women are increasingly left wing (Paywall, The FT).

Flowers of Fire by Hawon Jung describes the culture of sexism in South Korea today. Alice Evans has written a comprehensive review on her Substack.

1. Around 50,000 people die from venomous snakebites every year in India alone, but existing antivenom technology needs cold-chain storage and is made through a century-old method: by physically 'milking' snakes, injecting the venom into a large animal, and then bleeding that animal to harvest its antivenom. Recently, computational biologists at the University of Washington used an AI tool, called RFdiffusion, to create a protein that neutralizes a wide range of lethal snake toxins. The designed protein is stable at higher temperatures, so it doesn’t require refrigeration, and blocks toxins from latching onto neuromuscular targets. In trials it has protected mice from otherwise lethal doses of snake venoms.

   

2. In 2019, the first vaccine against the Ebola virus was approved for adults by the European Union and the United States. Although the vaccine was highly effective in protecting people who received it in clinical trials, it’s given out with a different strategy in the real world, using 'ring vaccination': this is when contacts or high-risk groups are vaccinated, rather than everyone in the population. This strategy can help contain an outbreak quickly, but it’s unclear how effective it is in the real-world. For a new study, researchers estimated that the vaccine reduced the chances of Ebola virus disease by 84 percent in the Democratic Republic of the Congo, which is highly effective. They arrived at this value using a test-negative design; when patients are tested for an illness like Ebola, in this case (cases test positive, and controls test negative). This method helps reduce biased results that could come from some groups being less likely to get tested than others.

3. Fears of a new flu pandemic have been growing, as bird flu outbreaks have occasionally led to human infections. In the US, there have been 67 confirmed cases in humans, but clinical testing is limited. So far, the virus doesn’t seem to have gained the ability to transmit between people, which means most infections have been from incidental exposure to infected wild birds, cattle, and contaminated raw milk. This might be why many cases have been mild so far and have not spread to the respiratory tract where flu infections are most dangerous. A new study shows that 93 percent of cases have had eye infections – specifically, conjunctivitis (or 'pink eye').

4. A large study, performed on 960 female mice, suggests that genetics – and not diet or exercise – are the biggest predictor of which mice live longer than others.

5. In ovo sexing technology has finally entered the US market, according to nonprofit Innovate Animal Ag. Something like six billion male chicks are killed in the egg industry each year. Farmers don’t have a use for them (and they don’t grow large enough, quickly enough, to be profitable for meat.) In ovo sexing technology could help identify the sex of a chicken egg before it hatches using hyperspectral cameras or polymerase chain reaction, which would help avoid culling live chicks. These technologies have been used in Europe for several years; Germany and France enacted bans on chick culling in 2022.

   

6. More than 500,000 Americans have lung cancer at any given time. Non-small cell lung cancer is the most common type, accounting for 80–85 percent of cases, and of them, about 10–15 percent have mutations in a gene called EGFR, which drives faster tumour growth. The mutation is much more common in Asia and lung cancer patients with it tend to lack a history of smoking and have more severe disease. Recently, a drug called Tagrisso has shown high efficacy against both early and late stage EGFR lung cancers. But new data from a phase 3 clinical trial, called MARIPOSA, shows an even greater efficacy from a combination of two drugs, called amivantamab and lazertinib, which the trial finds can extend their survival by an additional year, or more, over Tagrisso. The European Commission has already approved the treatment.

7. Purdue Pharma and the Sackler family reached a 7.4 billion dollar settlement with 15 US states to resolve lawsuits around their role in the opioid crisis, STAT reports. Their aggressive marketing of OxyContin, downplaying of its addictive properties, and influence over the FDA feature in the popular book Empire of Pain.

8. A second pig-to-human heart transplant has failed, according to a report in Nature Medicine. Surgeons at the University of Maryland extracted a heart from a gene-edited pig and transplanted it into a 58-year-old man with advanced heart failure. A total of ten genes were modified or added to the pig; four genes encoding sugars that can trigger human immune responses were deleted, and six human genes were added in (they help with stuff like inflammation and blood clotting.) The man did well for a few weeks, but then the heart stiffened up and failed. He died on day 40 post-surgery.

9. A gene-editing technology called 'dual prime editing' was used in plants for the first time. This tool can precisely delete up to two million bases of DNA, or replace a 258,000 base stretch of DNA with a new sequence, in both wheat and tomatoes (so far). The technology uses a special ‘fusion’ protein (essentially a modified Cas9 and reverse transcriptase enzyme tethered together) and two guide RNA molecules to make edits. The guide RNAs direct the protein to ‘nick’ two different sides of a DNA strand, opening up space to either paste in new DNA, invert the DNA, or delete everything between the nicks.

10. Scientists sifted through 2,000 microbial enzymes to find two that break down polyethylene terephthalate (PET) more quickly than any 'known' enzymes. The two PETases are called Mipa-P and Kubu-P. After their discovery, scientists engineered both enzymes to work under industrial-like conditions: an engineered form of Kubu-P, for instance, can break down 90 percent of PET in under 8 hours. Still, enzymes like this probably won’t be used in commercial recycling plants any time soon because they must contend with the rigid crystallinity of post-consumer plastics.

11. In May 2024, about 12 percent of American adults said, in a KFF poll, that they had used or are currently using GLP-1 drugs for diabetes or weight loss. Large-scale trials show that GLP-1 drugs minimize the progression of pre-diabetes, treat diabetes and kidney disease, reduce the chances of cardiovascular events like heart attacks and stroke, and reduce breathing problems during sleep. They also reduce overall death rates among people with cardiovascular disease. As more trials are run, we may find even more benefits of these drugs.

12. Chinese biotechnology companies now have 6,280 drugs in development – a 1,200 percent increase over one decade ago, and about two-thirds as many drugs in development as in the US – and are consistently beating Western companies to clinical trials. An American company called Pheast, for example, raised $76 million to make drugs targeting aggressive forms of breast cancer after its founder, Stanford professor Irv Weissman, discovered a new way to target such cancers in 2019. But Pheast was beaten to clinical trials by a company in China, called Antengene, after the latter 'no­ticed Weiss­man’s pa­per' and quickly developed a similar drug, according to reporting in Endpoints News.

    

13. OpenAI and Retro Biosciences (a longevity company funded by Sam Altman) say they’ve trained an AI model that can design transcription factors that are highly capable at converting regular cells into stem cells, a type of cell that can differentiate into many different cell types. This is typically done by ‘reprogramming’ a normal cell back into a stem cell using four different proteins, called Yamanaka factors. But this method doesn’t always work in all kinds of cells, so scientists have long wanted to find better factors. These AI-designed variants are apparently 50-times more effective than the normal Yamanaka factors (at what, exactly? None of the reports say), but neither company has yet released more complete data.

This issue of Links in Progress is dedicated to practicing and aspiring architects, developers, planners, policymakers, public officials, and the general populace, who are interested in creating a better built environment in their own communities. Enjoy!

Ablitt Tower

• Location: Santa Barbara, California

• Year Built: 2006

• Building Type: Single Family Home

• Unit Breakdown: Single Family Home

• Architect: Jeff Shelton

• Developer: The Ablitt Family

Ablitt Tower is a single family home in downtown Santa Barbara, California. Located on a tiny 20 foot x 20 foot lot, it demonstrates how whimsy and wonder can be imbued into the built environment anew. In its rejection of the standard front-yard-white-picket-fence-two-car-garage typology, it forces observers to question what a single family home can be. Not dissimilar to Gaudi’s impact in Barcelona, Shelton has delivered magical projects around the city, and has adorned its streets with tiles, lampposts, and railings.

Catfiddle Street

• Location: Charleston, South Carolina

• Year Built: 2007 - 2024

• Building Type: Inner block neighborhood

• Unit Breakdown: 24 residential lots, 30 plus units (several ADUs and apartments scattered throughout the lot)

• Architects: New World Byzantine, Craft Design Studio, American Vernacular, Liberatos Architects, Randolph Martz

• Developers: George Holt, Cheryl Holt, Jerry Moran, Reid Burgess, Sally Eisenberg, Vince Graham

Catfiddle Street is that exceptionally rare project that wasn’t developed all at once, but rather was carefully put together over nearly two decades. Each new home rose organically in consideration of its neighbors, resulting in a feeling that, much like the rest of Charleston, makes it seem like it has existed for centuries, and naturally belongs exactly where it is. Few places built in the last several decades can claim as much. Accessing Catfiddle is like stumbling upon a cherished secret, as its entrances are unassuming narrow pathways that squeeze in between homes that are hardly 10 feet apart. It is a marvel of ‘inner-block urbanism’, where development occurs within the block, tucked in rear yards, as opposed to traditional streetfront urbanism.

Albina Yard

• Location: Portland, Oregon

• Year Built: 2016

• Building Type: Mixed-Use Office

• Unit Breakdown: 16,000 square foot building with a ground floor cafe

• Architect: Lever Architects

• Developer: reworks

Offices are seldom integrated within a community. Either located in segregated ‘parks’ or cloistered in a Central Business District (CBD), their scale and anonymity mean that they rarely make for good neighbors. What makes Albina Yard so interesting is that it embodies all that most offices are not: rising four stories with handsome wooden window frames that pop out from the large glass windows, it respects its neighbors without compromising on the boldness of its design. Cross-Laminated Timber (CLT) panels provide a warmth that is almost entirely absent in the sterile world of offices-for-rent. Albina Yard is an excellent precedent for suburban neighborhood commercial buildings.

Wellington House

• Location: Wimbledon, London

• Year Built: 2020

• Building Type: Mixed-Use Office

• Unit Breakdown: 48,000 square foot speculative office building with shops on the ground floor

• Architect: MATT Architecture

• Developer: Columbia Threadneedle

A wonderful transformation of a dreary old structure, Wellington House is a delightful mixed-use office building. Honoring the corner with graceful rounded windows, the usage of glazed ceramic tiles and sawtooth brickwork elevate what would otherwise be a nice-enough building to an exceptional one, whose gravity pulls passersby in to more closely inspect its thoughtful detailing.

Berkeley 6

• Location: Portland, Oregon

• Year Built: 2019

• Building Type: Townhomes

• Unit Breakdown: 6 homes

• Architect: Polyphon

• Developer: Ethan Beck Homes

Residents of suburban neighborhoods often express fears about new non-single-family housing changing the character of their communities. While these concerns may or may not be made in good faith, there are many wonderful new projects that prove these fears are unfounded. From the street, Berkeley 6 looks like a single family home with a two-gabled elevation. From the side, it reveals itself to be 6 townhomes, perfectly in keeping with the character of the neighborhood. For suburban areas that want to add more housing without feeling as though the essence of the neighborhood is changing, Berkeley 6 proves this can be readily accomplished.

E. Bronson Ingram College at Vanderbilt University

• Location: Nashville, Tennessee

• Year Built: 2018

• Building Type: Residential College

• Unit Breakdown: 350 beds, a dining room, great room, and series of lounges throughout the building

• Architect: David M. Schwarz Architects, HASTINGS Architecture

• Developer: Vanderbilt University

Vanderbilt University in Nashville, Tennessee has transformed its campus over the last two decades from a nice-but-unremarkable place, to one of the most exceptional locations for higher learning on the continent. Accommodating 5,200 students in a series of residential colleges, it has proven that one of the best ways to attract students is to be attractive.

Pauli Murray College & Benjamin Franklin College at Yale University

• Location: New Haven, Connecticut

• Year Built: 2017

• Building Type: Residential College

• Unit Breakdown: 500 beds, a dining room, library, coffee shop, gym, theater, pottery studio and series of lounges throughout the buildings

• Architect: Robert A.M. Stern Architects

• Developer: Yale University

Drawing on James Gamble Rogers' original residential colleges in New Haven, Yale has recently completed some of the country’s most extraordinary new buildings. Housing 904 students across 532,000 square feet encompassing dining halls, lounges, study rooms, classrooms, and multi-purpose space, the project is estimated to have cost between $500 million to $600 million in 2013 dollars. Though not cheap, it is a testament to America’s ability to create exceptional places anew at the confluence of ambition, prosperity, and skill.

Chevy Chase Lake

• Location: Chevy Chase, Maryland

• Year Built: 2023

• Building Type: Mixed-Use Apartment Complex

• Unit Breakdown: 530 housing units (a mix of condos and rentals) and 110,000 square feet of retail, including a grocery store

• Architect: David M. Schwarz Architects

• Developer: The Bozzuto Group and The Chevy Chase Land Company

Located adjacent to a Purple Line stop on the DC Metro, Chevy Chase Lake is likely the finest collection of new mid-rise buildings in the country. A sterling example of transit-oriented-development made all the more remarkable given it is located in a highly affluent, and historically anti-development, suburb.

The Fitzroy NYC

• Location: New York, New York

• Year Built: 2019

• Building Type: Condo Building

• Unit Breakdown: 14 homes

• Architect: Roman and Williams

• Developer: JDS

Terra cotta has been revived in gloriously wonderful ways in New York in the last decade. The ultimate example of this has been The Fitzroy, the luxury condo building located just a few steps from the High Line. Shimmering in the daytime when light hits its glazed facade, and brooding enchantingly in jazz-age sophistication at night, it proves that height is not a deterrent to good design nor compelling city-building. Indeed, there are few places that wouldn’t benefit from the construction of similar buildings, either lording as landmarks over their lower-rise dominion, or demanding attention in a sea of painfully anodyne mid-rise structures.

Second Home Hollywood

• Location: Los Angeles, California

• Year Built: 2019

• Building Type: Office Complex

• Unit Breakdown: 60 garden office pods

• Architect: SelgasCano

• Developer: Second Home

Second Home is a breath of fresh air amidst the highways and parking lots that populate Los Angeles–a true forest amidst the concrete jungle. Acknowledging that the city’s climate is its most valuable asset, the colorful pods that make up this office complex blur the lines between the indoors and the outdoors, with 360 degree glass walls and workspaces in the gardens that allow tenants to take full advantage of the California sun.

Weespergilde

• Location: Weesp, North Holland

• Year Built: 2013

• Building Type: Single Family Home Community

• Unit Breakdown: 42 Homes

• Architect: Braaksma & Roos

• Developer: HBB

Weespergilde breaks the mold of suburban development through its bold use of natural materials in quasi-vernacular architecture. Part of what makes this project so compelling is that it’s familiar, with a slight twist. Designed for families and oriented around courtyards, it has far higher density than most North American subdivisions, but remains eminently genial, attractive, and comfortable.

Spoke Street Shophomes

• Location: Wheeler District, Oklahoma City

• Year Built: 2020

• Building Type: Live-work Townhomes

• Unit Breakdown: 10 Homes, a mix of commercial, retail, and residential ground floors

• Architect: Sam Day (Dryline Architecture)

• Developer: Sam Day

At 15 feet wide, three stories tall, and with flexible ground floor space, these 10 live/work townhomes are wholly novel, yet totally alluring. Rarely are new townhomes so fine grained in form. Rarer still are when any use is permitted on the ground floor (commercial, retail, workshop, or residential). The result in an intriguing mosaic that takes on the disparate personalities of the owners of each building, while the overall design language remains intact.

Bonte Cafe in Hoi An

• Location: Hội An, Vietnam

• Year Built: 2021

• Building Type: Cafe

• Unit Breakdown: A three-story cafe

• Architect: Yên Architecture

• Developer: Bonte Cafe

Concrete can be a cold, austere, and forbidding building material when used carelessly. But when it’s tactfully designed and married with greenery, the resulting structures can be very elegant. And when the building leans into the public realm, as with this cafe, there are few places one would rather spend their time.

The Museum of Meenakari Heritage and Sunita Shekhawat Store

• Location: Jaipur, India

• Year Built: 2024

• Building Type: Mixed-Use Structure

• Unit Breakdown: A museum and a jewelry store

• Architect: Studio Lotus

When I first came across this building, I didn’t think it was real. The level of detail, and ambition of design, are all but nonexistent in contemporary building practices. Upon consulting Google Street View, and seeing construction as well as completion images, I was not only wonderfully surprised, but inspired. The Jodhpur red sandstone is an exquisite building material, honored perfectly by the latticework ornamentation, making what might otherwise appear a squat structure grand and dignified.

Fort Greene Five

• Location: Brooklyn, New York

• Year Built: 2023

• Building Type: Townhomes

• Unit Breakdown: 5 homes

• Architect: Boro Architects

There exists a popular sentiment that we can’t build as well as we used to. That our best days are behind us. These new townhomes in Brooklyn, thankfully, should disabuse many from this notion. These Brooklyn townhomes are indistinguishable from their predecessors that rose a century and more before. We should not always mimic styles from the past, but where context and will demand it, it is simply not true that we cannot build as well as we once did. Indeed, in some respects, we can build better.

1. The global relationship recession: a rise in the number of single people is becoming a key driver of falling birth rates. Policymakers who want to see birth rates rise should consider how they can help people form relationships.

2. More evidence that access to housing is a fertility bottleneck. A new paper has found that receiving money through a lottery can boost the probability of having a child by 32 percent among Brazilians aged 20-25, and by 10 percent for those aged 25-35. But Lyman Stone, on his new substack, explains that these housing lotteries are not lotteries as we traditionally understand them but instead usually a way of extending credit to people who cannot get mortgages. Not only are the participants demographically unusual, they are also stuck paying in for years before they ever receive one. There’s a good chance that participants merely delay their fertility until they win a house.

3. For the first time since 2015, South Korea has recorded an increase in annual births – up by 3.1 percent between 2023 and 2024. The government’s pro-parent policy push might have contributed to this rebound, but it could just be a ‘dead cat bounce’: the result of a mini-boom of babies born in the 1990s reaching their early 30s.

4. Falling birth rates mean advanced economies could see per capita economic growth slow by 0.4 percent per year from 2030 to 2050. Consultancy giant McKinsey publishes new analysis on how the global birth rate challenge will pressure public finances and depress living standards all over the world.

5. Should we give children the vote? The authors of this working paper suggest that enfranchising children would increase the birth rate. They assume that children, “as perfectly informed and rational utility-maximising voters” will vote to redistribute state spending away from pensions and towards families with children, thus reducing the costs of having children.

6. Married Chinese women are being cold-called by local government officials asking about their plans to have children. The FT reports on China’s escalating efforts to raise birth rates.

7. Building your own ‘village’: you can control exactly how your kids are raised, or ask for community help with child care – but you can’t have both. In The Atlantic, Stephanie H. Murray writes about the benefits and compromises she discovered in creating an informal childcare network, or 'baby swap', with four other families.

8. India is getting old before it gets rich. Many parts of India have birth rates on par with western countries, and are rapidly aging and dealing with shrinking workforces. States including Tamil Nadu and Andhra Pradesh are considering policies to incentivize citizens to have more children

9. Biotech company Gameto has developed a process to mature eggs outside the body and announced the birth of the first baby born via the technology. Gameto’s new process reduces by 80 percent the number of hormone injections women undergoing IVF are required to take and makes the treatment shorter

10. Superstar economist Claudia Goldin argues that the rapid economic growth experienced by countries like South Korea and Japan has led to a clash between traditional expectations upon women and their new economic opportunities that means they then have fewer children. Goldin does not empirically test her theory.

11. Listen to Ezra Klein respond to a question about choosing parenthood. Klein encourages people not to think about becoming a parent only in personal pro and con terms:

It’s amazing that my children get to experience life, not because it makes my life better in every respect, but because their lives are precious things…the joy of parenting is not the joy of parenting but the joy of your child’s existence.

Klein also cites Gilead by Marilynne Robinson (perhaps the best novel I read in 2024) for its portrayal of parent joy. This short book is about an Iowan preacher who unexpectedly becomes a father in later life.

12. Each cohort of young Europeans is delaying the major life events of early adulthood – like getting their first job, leaving home, moving in with a partner, and having their first child – by more and more each generation. Those with highly educated parents are most likely to postpone all these life events more.

13. Don’t turn birth rates into a culture war issue. In an article published in the summer, Addison Del Mastro argues that attempts to subsume the birth rate challenge into culture war discourse distract from policies that can actually help more people start families and will never convince anyone unsure about parenthood to take the plunge.

14. The Ancient Greek historian Polybius (c.200-118BC) blamed falling births in his own time on young men becoming distracted by the 'pleasures of an idle life' and the desire to bring children up in 'extravagant luxury'.

Planes

1. A new airport has opened in Greenland’s capital of Nuuk. With a longer runway (2,200 meters, replacing the older 950-meter runway), larger planes that previously had to use former US military bases on the territory will now be able to fly directly. Direct flights from New Jersey are launching in June, making the territory even more appealing to those searching for land to create a new city-state.

Trains

2. Vietnam has granted approval to a new $67 billion high speed line between Hanoi and Ho Chi Minh City (Saigon). The line will be one of the most ambitious high speed projects outside of China, covering a total distance of 1,541 kilometers (957 miles: for comparison, California High Speed Rail Phase 1 is ~500 miles) at a cost of ~70 million USD per mile. Construction is planned to commence in 2027, with the first services to run in 2035.

3. The UK’s High Speed 2 rail project spent £100 million GBP ($127 million USD) on a one-kilometre-long steel mesh over a section of the line to protect nearby bats. The chairman of the project also revealed that 8,276 separate consents were required in order to proceed with construction of the line, which runs from London to Birmingham. It turns out that the bats aren’t even particularly rare.

4. High speed rail may also be coming to Australia: drilling has begun on the New South Wales coast to identify the best routing for an initial line, planned to run between Sydney and Newcastle. Construction contracts for the first stage will launch in 2027 with services commencing around a decade later. The Australian Government has also approved the next stage of the existing Inland Rail project, with the goal of building a dedicated freight railway between Melbourne and Brisbane.

5. Ghana has opened a new railway running from Tema to Mpakadan, as part of an eventual 1,200-kilometer line that will link Ouagadougou, the capital of Burkina Faso, with the Ghanaian port of Tema. Built at a cost of $447 million USD, the 97-kilometer (~60 miles) stretch comes to around $7.5 million USD per mile and will support freight movements between inland territories and the coast. The Ghanaian government plans to develop around 4,000 km of railways by 2030.

6. The first three lines of Riyadh’s new metro network opened on the 1st December, with another two following on the 15th December. The sixth line of the initial network will open in early January. The six automated metro lines, with a total length of 176 kilometres (~110 miles) and 85 stations, have been constructed in just a decade at a cost of $22.5 billion USD (~$205 million per mile). The project has included some novel financing methods, including the sale of station naming rights creating destinations such as the ‘SAB Bank’ and ‘STC’ (after the Saudi Telecom Company) stations.

7. The French government has announced a plan to build 22 new intermodal rail freight terminals (where containers can be moved between trains and lorries) across the country, as well as expanding several existing ones. The strategy and new investments will contribute to the French’s Government’s 2019 ambition to double rail’s share of the freight market to 18% by 2030.

8. Switzerland and Germany have agreed to jointly invest in upgrading a cross-border rail line to relieve congestion on existing routes, and decrease journey times. The 430 million CHF investment will electrify the 75 kilometre (~47 miles, ~$10 million USD per mile) route between Basel and Erzingen, as well as extending and raising platforms, installing noise barriers, and renewing electrical systems to support a more intensive rail service.

9. In the twelve years from 1995, Madrid tripled the size of its metro at one of the lowest per-km construction costs in the developed world, thanks to a unique model of infrastructure planning, financing, and construction. Ben Hopkinson wrote for Works in Progress about what lessons can be learned from the Madrid-method for other cities looking to build at scale.

Automobiles

10. After years of wrangling, New York City’s road congestion pricing will finally go live on Manhattan’s highways on January 5th. The new $9 a day toll is less than the previously planned $15 charge, and revenues will be used to support the Metropolitan Transportation Authority to deliver its capital plan. New York State Law requires around $15 billion to be raised from tolls towards the capital plan, and the charges are expected to climb to $12 a day in 2028 before reaching $15 in 2031.

Boats

11. Cambodia is building a 180-kilometer (112-mile) canal connecting the Mekong Delta in the North with its southern coast, enabling ships and freighters to navigate from the capital to the sea without traversing through neighbouring Vietnam. At a price of $1.7 billion (~$15 million per mile, ~a fifth of the cost per mile of France’s new canal), the Funan Techo Canal’s cost is equivalent to around four percent of the country’s GDP. However, the project has faced concerns with funding, and of environmental effects on the already-fragile Mekong water system.

12. Operational tests have begun on India’s New Pamban Bridge, the first vertical lift sea bridge on the subcontinent, where a horizontal bridge span rises vertically, severing the bridge, but allowing boats to pass underneath. The new bridge and the railway it carries cost 535 crore INR (~$64 million USD) for its 2.05km length (~$49 million USD per mile) and a lifting span that gives ships a clearance of 22m above sea level. The project has not been without controversy, with accusations of safety lapses throughout construction.

Everything else

13. Monaco, one of the smallest territories in the world, is reclaiming land from the sea to create a new district. This is expected to cost around two billion euro: half for the reclamation, and half for the development that sits on top of it. The six-hectare new district of Mareterra with its 110 apartments, 10 houses, and 4 villas (costing around $200 million each) will grow Monaco’s area by around three percent. With prices per square metre of up to €120,000 EUR (more than $12,500 USD per square foot), Mareterra will be home to some of the most expensive real estate in the world.

14. In Athens, the regeneration of the former Ellinikon airport site, part of which temporarily supported the 2004 Summer Olympics, is delivering 2500 hectares of new urban space and 240 acres of parkland. The site will be home to around 30,000 people, including the Riviera Tower, a 170 home tower that at 198 meters tall is almost double the height of Greece’s current record holder. The project is an example for other cities looking to maximise the utility of inner city land by moving industrial and logistical sites further out.

15. The UK’s Department for Energy and Net Zero and Westminster City Council have jointly announced plans for a new district heating network in the heart of the British capital. With up to £1 billion GBP ($1.27 billion USD) to be invested between now and 2050, the ‘South Westminster Area Network’ will be one of the largest in the UK, with ambitions to not only service homes and businesses, but also key landmarks including the Houses of Parliament (built in the 1800s) and the National Gallery. Heat networks can provide hot water and heating 40% cheaper than heat pumps can by generating heat at scale and using waste heat from sources such as the River Thames and the London Underground. An article in Works in Progress explained how they work.

16. Developers in San Francisco (home to some of the highest land values in the US) are turning to slimmer building designs to take advantage of looser building regulations. Buildings of up to 73 metres in height (240 feet) are being proposed on single lots of just 7.5-10.5 metres width (25-35 feet). The greater development permitted on single lots supports home construction by reducing the need for developers to assemble larger plots. These developments are coming forwards as San Francisco enters the final stages of a project to rewrite zoning rules.

‘Links in Progress’ newsletters are semi-regular roundups of interesting stuff that's happening in topics that we care about. In this one, Boom’s Phoebe Arslanagić-Little reviews the most important things happening in the world of pronatalism and family policy. You can opt out of Links in Progress here.

1. China’s population will shrink by 51 million people by 2035, Bloomberg Intelligence analysts predict (clip). That’s more than the population of Canada. 2035 is also the year that China’s main state pension fund is predicted to run out of money.

2. Chinese President Xi Jinping sets out his plan to deal with China’s demographic challenges (in Chinese, here is a summary translation). This comes after a new directive (in Chinese) from China’s State Council outlined 13 policies to make China more family-friendly, including better maternity leave and childcare.

3. Half of US counties have lost population between 2010 and 2020 – because of falling birth rates, rather than changes in net outward migration, this new working paper finds.

4. Official forecasts of what America’s social security will cost are being made on the assumption that US birth rates will increase to 1.9 children per woman by 2040. This would be a return to 2011 fertility levels. The OASDI Trustees Board is the statutory body overseeing the financial integrity of America’s Social Security trust funds.

5. Telenovelas may have helped create a habit of having smaller families in Brazil. Telenovelas – TV soap operas – often depict small families, in part because it makes the plot more manageable for showrunners and writers. This paper found that the more Brazilian women were exposed to telenovelas, the fewer children they had.

6. A feel-good watch for over the holidays. Check out the trailer for JOY – The Birth of IVF, a new film starring Bill Nighy and James Norton, about the scientists who developed IVF.

7. Can lack of access to contraception explain high fertility in sub-Saharan Africa? Perhaps not. This 2024 working paper finds that providing vouchers covering 100% of the cost of contraception to 14,000 households in Burkina Faso for 3 years did not reduce the number of children being born or drive take-up of modern contraceptives.

8. Missing cousins. This paper projects that a 65 year old woman living in South America or the Caribbean in 1950 could have expected to have 56 living relatives in her extended family (great grandparents, grandparents, parents, siblings, aunts, uncles, nieces, nephews, and first cousins), but only 18 such relatives in 2095.

9. Farewell to the Highlands – Scotland will start depopulating in under a decade. Scotland’s birth rate is 1.28, closing in on Japan’s 1.2. The Scottish Government’s action plan focuses on how to help rapidly depopulating parts of Scotland cope with the draining of their communities.

10. Better local institutions might make it easier for mothers to balance career and parenthood. A new study looking at European regions finds a one percent improvement in local institutional quality – using the European Quality of Government Index which includes factors like quality of public services – correlates with an eight percent increase in local fertility rates.

11. Are smaller families driving a stronger preference for sons over daughters in India? Northern Indian respondents’ expressed desire for at least one son, particularly an eldest son, increases as hypothetical family size decreases.

12. Is the rise in singles a major factor driving global birth decline? So argues social scientist Alice Evans. In the EU, twenty percent of men aged 25-54 live alone and eleven percent of women. Sweden leads the pack: fifty percent of all Swedish households consist of single people.

13. Finally, a reminder that anti-natalist sentiment is no modern phenomenon. Lyric poet Theognis of Megara writes that it is better never to have been born, in approximately 600BC:

1. The UN’s population predictions and birth rate forecasts keep getting it wrong – there were 30% fewer births in Colombia in 2023 than the UN recently predicted. John Burn-Murdoch explains why peak population might arrive sooner than we think.

2. Ask not what you can do for your baby, but what your baby can do for you. What, exactly, is the point of having children? Philosophers Anastasia Berg and Rachel Wiseman investigate in their new book, What are Children For?: On Ambivalence and Choice.

3. 22% of Americans say not enough babies are being born in the US, but 23% say too many are. YouGov asked the same question of Brits last month and the results are strikingly similar: 20% of Brits say too many children are being born in the UK, and 21% say not enough are. Among Americans, groups more likely to say too many American children are being born include liberals, younger adults, women, and non-parents.

4. In most human societies, poverty does not predict higher fertility, argues US demographer Lyman Stone. He argues the relationship between income and fertility is much more complicated than often claimed, and cultural differences play a big role.

5. In fact, a new study finds that wealthier Dutch people are increasingly more likely to become parents than those on lower incomes. Author Daniël van Wijk also published work earlier this year identifying an intensifying relationship between income level and chance of becoming a parent across multiple countries, including the US. 

6. Philosopher Victor Kumar makes the case for progressive pronatalism in the NYT – he argues that if you care about the poor, the sick, and the old, then you should care about falling birth rates too.

7. On the Japanese island of Tokunoshima, fertility has been above the national average for years. Tokunoshima is home to Japan’s two highest fertility municipalities, with respective Total Fertility Rates of 2.25 and 2.24, in comparison with 1.2 for Japan overall. 45% of women aged 20-24 are married in Tokunoshima, in comparison with 7% nationally, and that getting married after falling pregnant is normal and not stigmatized – that could be an important part of the story behind the island’s higher fertility. 

8. Taxing the stork? This paper argues that low-fertility countries are not doing enough to take into account and alleviate the costs of raising children upon parents. Even in Nordic nations where parents receive more state support than in other countries, that support does not fully reflect the costs parents face, and parents in these countries (most of whom work) have bigger tax bills. There’s a shorter write-up by study co-author Pieter Vanhuysse here. 

9. Low fertility is bad for economic growth, says economist Jesús Fernández-Villaverde in an interview with social scientist Alice Evans. Watch on Youtube or listen on Spotify. Evans has also written about different explanations for the global decline in fertility and the need for a response to this challenge.

10. Economist Emily Oster, made famous by her data-driven approach to pregnancy, has a parenting podcast with The Free Press. The latest episode features Ross Douthat and Bryan Caplan making the case for having children and the fun of being a parent. 

11. Ultra-low fertility South Korea has awarded civilian service medals to two mothers for each having 13 children. 59-year old Mrs Lee used her moment in the sun to tell journalists that South Korea “desperately needs” a more parent-friendly workplace culture. 

12. A record high of 4.7% of South Korean babies were born outside marriage in 2023, in comparison with only 2.1% in 2013. For comparison: just under 40% of US babies are born outside marriage, 51% of babies in England and Wales, and 65% of French babies. 

13. Research in the US suggests that home working may have a particularly positive effect on some groups of women. Specifically,  women who are doing well financially, and older women who already have children.

14. 40 years after Christ’s death, there were perhaps 1000 Christians, but by 400 AD there were 40 million of them. Scott Alexander reviews The Rise of Christianity, which investigates what factors may have driven the Christian triumph, including the theory that higher fertility was a major contributor.

15. Different strokes for different folks. A paper looking at family policies in the UK, Germany, and Finland – including more generous paternity leave – shows how and whether such policies affect childbearing varies significantly by women’s education level, and also between countries.  

16. While Scotland is seeing a rise in one-child families, the two-child family model remains the norm in England and Wales. 

17. How many children would you like? This paper found that to have a 90% chance of having 3 children without IVF, a woman should start trying for her first child aged 23. And for a 90% chance of having 2 children, she should begin at age 27. Demographer Anna Rotkirch has created the below visualization of the paper’s conclusions.  

18. Forward-looking thoughts. Finally, I recently read George Eliot’s 1861 novel Silas Marner, about the transformative effect of an adopted child on the life of a recluse whose only comfort has hitherto been money:

“Unlike the gold which needed nothing, and must be worshiped in close-locked solitude… Eppie was a creature of endless claims and ever-growing desires, seeking and loving sunshine, and living sounds, and living movements… The gold had kept his thoughts in an ever-repeated circle, leading to nothing beyond itself; but Eppie was an object compacted of changes and hopes that forced his thoughts onward…”