More recently, Martin Gould countered that you can’t hold demand constant when the whole reason demand grew is that efficiency drove prices down. The growth in chicken consumption might itself be a Jevons paradox effect.

Of course Martin is right that I was being hand-wavy in order to point out that the “efficiency = bad” meme is oversimplified and counterproductive. But it’s equally as hand-wavy to observe chicken price going down, chicken production going up, and assume it’s all a Jevons effect.

So I built a model to tease out what’s actually going on. Doing so reinforced my thesis that increasing efficiency has historically been the best way to decrease the number of animals needed for agriculture. But that doesn’t mean that efficiency is always good — the data tells a more nuanced story. If we shift analysis to the perspective of aggregate flock welfare, rather than flock population size, then we can see some of the potential downsides. However, seeing the downsides in this way also clarifies why the techno-optimist perspective is a powerful way to approach farm animal welfare.

How Things Have Changed Since 1960

The first question we’d like to answer is how many fewer chickens do we now need due to efficiency improvements? Previously, my oversimplified method was to look at how chicken size and growth rate have changed, then determine how many chickens-of-the-past would be needed to meet today’s demand.

However, in reality the chicken-of-the-past would be more expensive than today’s chicken, meaning people would eat less of it. How much less is a question of the price elasticity of demand: the increased percentage of chicken that people would consume given a 1% drop in price.

Once we know the price elasticity of demand, then we can more directly answer the question. Since wholesale chicken price is a reasonable proxy for efficiency, we can choose a point in the past, look at the growth rates, size, and price of chicken during that time, use the elasticity to figure out how much chicken would be consumed if it were at that price, then figure out how many chickens at that efficiency would be needed to meet that demand.

The true elasticity of chicken is an economic abstraction that varies significantly by context. However, there have been numerous studies on the topic that give a sense of the range, and it’s sensible to think the aggregate price elasticity since 1960 would fall in that range. The most recent estimate (Luke 2026) puts it between -0.6 and -1.1 (meaning that a 1% decrease in chicken prices leads to a 0.6-1.1% increase in chicken consumption). An older review (Andreyeva 2010) gives -0.44 to -0.92. Lusk 2016 finds larger elasticities up to -2, but using hypothetical purchase scenarios rather than actual transaction data, where price is more salient to respondents than at the register.

In order to be comprehensive and conservative, I considered a distribution of possible elasticities, from -0.44 to -2, centered on -.85. I gave higher-magnitude elasticities less weight since they’re possible but less likely.

I then modeled a comparison between the chicken industry of 1960 versus today. The results are represented by the four charts above:

• The top left chart represents the range of elasticities considered and their respective likelihood.

• The top right chart represents how much of the increase in chicken demand is explained by prices coming down — the “Jevons paradox” channel. Compared to 1960, we eat almost seven times more chicken as a country (3.7 times as much per capita) and it costs almost 50% less. In our base case, lower prices only explain 28% of the observed demand, although the possible range is anything from 14-65%. The rest is explained by population growth, income growth, and a secular increase in demand for chicken (e.g., a health-driven pivot away from red meat, product innovation like boneless skinless breasts, etc.).

• The bottom left chart represents the difference in the number of chickens we need today because of increase in efficiency. In the baseline case, we need 1.2 billion fewer chickens – a 14% decline. That said, if elasticity were at the higher end of the range, we would need more chickens, but 65% of the total probability mass is in the green (and the 1995 analysis below paints a clearer picture).

• The bottom right chart represents the difference in chicken-days, to capture the shorter amount of time that animals currently spend on farm. In the base case, 229 billion fewer chicken days are needed to meet current demand, although there are some unlikely scenarios in which more chicken days would be needed.

When is less actually more?

This model is an abstraction over a much more complex reality. But there are also more intuitive ways to understand why Jevons paradox doesn’t apply as much to animal proteins.

Firstly, it’s important to understand that a huge part of what consumers pay for animal proteins is retailer margin — these days it’s over half (55%) for chicken. This means that when efficiency changes wholesale price, the signal is heavily diluted by the time it reaches the end consumer.

Let’s consider what this means for a context where we increase efficiency through lowering mortality. A -2 elasticity (the largest found in the literature) means that every 1% drop in consumer price leads to 2% more chicken sold. On the surface, this seems damning for my perspective – if we increase efficiency by 1% by decreasing pre-slaughter mortality by one percentage point, then we might cause 2% more chicken to be produced!

But not so fast – a percentage point drop in pre-slaughter mortality is a 1% decrease in the price of wholesale chicken, which is what the retailers buy. Even if the retailers fully passed this on to the consumer (which they likely wouldn’t), it would translate to a 0.45% decrease in consumer prices, which under our conservative assumptions would increase consumption by 0.9%. And 0.9% more chicken consumed, at 1% better placement productivity, means fewer chicks need to be placed on farms.

This example pertains to mortality reduction, which is the kind of efficiency most aligned with welfare. But it illustrates how it can be possible to achieve both fewer chickens farmed and better welfare, even if total chickens consumed goes up.

Subscribe now

Secondly, let’s think about the canonical examples where Jevons paradox has more dramatic impact. One is the invention of the steam engine, which significantly increased the efficiency of converting coal into power. The resounding success of the steam engine paradoxically increased the total amount of coal consumed. Or, more recently, Jevons paradox has been a topic of discussion in the AI world because there could be an explosion of intelligence-derived work as AI decreases the cost of intelligence.

In these cases, the “paradox” comes from the fact that significant increases in efficiency open up completely new use cases for the resource in question. The steam engine created new demand for power generation in contexts where previously using coal would have been infeasible or uneconomical. AI is changing the fundamental dynamics of how intelligence is used in the economy.

Chicken doesn’t fit that pattern. A 10% drop in chicken price doesn’t open up some previously unreachable use case. In theory, one might expect a Jevons paradox effect if, for example, one kind of animal protein became suddenly cheaper than one of its substitutes (e.g. if pork suddenly became cheaper than chicken we might see demand for pork explode). But chicken is already the cheapest meat at the grocery store. There’s no substitute left to undercut.

Welfare Footprint

I think there’s an important steelman of the anti-efficiency perspective, which happens when we shift perspective to the welfare of individual animals.

The Welfare Footprint Institute is a research group that quantifies animal welfare by measuring hours spent at different intensities of pain, sorted into four categories: annoying, hurtful, disabling, and excruciating. They’ve identified the biggest welfare problems in poultry production and estimated how many hours of each kind of pain those problems generate.

A previous analysis they conducted directly compares the welfare of conventional broilers to broilers in a “reformed” scenario, meaning slower-growing breeds like the Hubbard JA757. The reformed birds experience fewer hours of excruciating, disabling, and hurtful pain than conventional broilers, though slightly more annoying pain because they live longer.

In order to integrate this into our existing model, we have to choose a different year rather than 1960. WFI’s reformed bird grows at about 45g/day, but the 1960 broiler grew even slower than that, which makes direct comparison difficult. However, 1995 was the year in which the conventional broiler of that time grew at a similar rate to today’s reformed birds. So if we run the analysis with efficiency held constant in 1995, then we can more directly compare the welfare of birds now versus then. The comparison isn’t perfect, since today’s reformed birds are purpose bred to have higher welfare, but it’s a start.

Under this analysis, efficiency gains since 1995 mean that we need 1.36 billion fewer chickens and 61.4 billion fewer chicken-days. The difference in chickens needed is positive rather than negative for almost the entire range of considered elasticities, making it different from the 1960 analysis. This is because the source of efficiency gains shifted over time. Between 1960 and 1995, most gains came from faster growth and better feed conversion; after 1995 they came almost entirely from bigger birds — today’s broiler is over 40% heavier than a 1995 one. Bird size is mainly what lowers headcount, since fewer animals are needed per pound of meat.

But looking at aggregate welfare tells a different story. Disabling pain — pain severe enough to interfere with normal function — has gone up by 231 billion hours, a 148% increase over what we’d see at 1995 productivity. Hurtful pain has gone up by 253 billion hours, an 11% increase.

So there might be reasons to worry about efficiency in animal agriculture after all, but they’re not the ones that Martin had in mind. Efficiency most likely decreases how many animals we need in animal agriculture – Jevons paradox doesn’t cleanly apply. But efficiency has also possibly made the welfare of individual animals worse.

At least that’s true at the current technological frontier.

If we further break down the WFI numbers into what’s causing the increased hours in pain, we see a number of discrete issues like lameness, chronic hunger in breeders, heat stress, and ascites. Each of these issues has technological solutions currently in development. Lameness can be mitigated through electron-beam inactivated vaccines, chronic hunger can be addressed through GLP1s or Layers Laying Broilers, ascites and heat stress can be mitigated through CoQ10, and so on.

These technologies are promising because they both improve welfare and raise efficiency — that’s what gives them a path to commercial adoption. If one starts from an anti-efficiency perspective, that might look like a strike against them. But this analysis hopefully shows why this is wrong. The net effect of efficiency has been historically mixed because the upsides (less agriculture) have to be traded off against the downsides (lower welfare per bird). But if we know a technology will directly solve a major welfare issue, then we aren’t faced with this tradeoff. We get less agriculture and better welfare — a Pareto improvement.

In the first essay I ever wrote on this blog, I distinguished between two broad approaches to externality mitigation: degrowth and techno-optimism. When abundance generates externalities, degrowth says sacrifice the abundance. Techno-optimism says innovate your way out of the tradeoff. This analysis gives a more precise way to understand this distinction.

If one thinks the problem with animal agriculture is “efficiency,” then the natural solution is to sacrifice efficiency, leading to degrowth. But once we see that efficiency itself reduces how much agriculture we need, and the actual problem is the downstream effects of efficiency on individual welfare, the question can instead becomes how to mitigate those effects directly. Understanding the nature of the problem ensures we’re pointed at the best solutions.

The problem current projects face is lack of a concrete value proposition. They hold the promise to generate massive amounts of data, but then don’t have a great answer to what to do with all that data – neither for the betterment of production efficiency nor welfare. Better AI over the next few years will help, but it won’t solve the fundamental problem that information is only as useful as the actions it enables.

PLF for poultry is struggling because there’s a missing piece of the puzzle that, to my knowledge, no one is yet addressing – individualized treatment. This is a hardware and logistics problem that, once solved, could give PLF in poultry the value proposition it’s been missing, and unlock a category of health interventions that flock-level economics has kept out of reach for as long as the industry has existed.

The Economics of Poultry Health

The fundamental challenge in poultry health is that interventions happen at the flock level. Given that there can be tens of thousands of birds in a single broiler house, a farmer usually can’t tell which specific birds are sick. And even if they could, there’s no way to treat just those birds. So when something needs treating, the drug goes to every bird in the barn, including the thousands that don’t need it. The per-bird drug cost gets multiplied across the whole flock, which puts massive economic pressure on every health intervention to be as cheap as possible.

PLF tools could, in theory, help the farmer detect and identify sick birds earlier (although simply tracking individual animals is still a frontier problem in poultry PLF, which some companies are currently trying to address). But it wouldn’t solve the brutal economics of poultry health management.

Suppose you have a condition that kills 0.1% of the chickens in your 40,000-bird barn (40 birds in total). At $4 of economic value per bird, the condition costs the farmer $160 per flock. To break even on a treatment, you’d have to dose all 40,000 birds for less than 0.4 cents apiece. Once you account for management overhead, manufacturing costs, and any uncertainty about efficacy, the real ceiling is lower still. Most likely, given current economics, this condition would go untreated, and the mortality would be considered a necessary if unfortunate cost of doing business.

Now suppose that you could treat only the birds affected by the condition. Suddenly the budget is $4 per bird, a 1,000x difference! This still might not sound like a lot, but pharmaceuticals often have significant pricing flexibility based on the cost-sensitivity of the target market. With this new economic equation, many existing drugs that are currently priced out of poultry come back into reach, and entirely new ones become worth developing.

The unlock here is for reactive treatment, not prevention. For something like a vaccine, where the point is to prevent a health issue from occurring, flock-level administration still makes sense. But for an intervention meant to treat an existing health issue that the farmer detects in certain birds, flock-level treatment inevitably means there will be some amount of waste.

Antibiotics are one potential use case, as the most common form of health treatment that birds currently get. Over the last few years, the industry has moved away from them in response to public health concerns about antimicrobial resistance. Flocks raised without antibiotics earn a “No Antibiotics Ever” label that carries a premium at retail. The goal is laudable, but the shift is one likely reason mortality in the US broiler sector has climbed to almost 6% in recent years. Clearly, farmers are in need of new tools to ensure positive broiler health.

There are many possible alternatives that can lower the risk of bacterial pathogens while not contributing to antimicrobial resistance. Bacteriophages, for example, are viruses that infect and kill a specific species of bacteria, then die off once their target is gone. Because the mechanism is completely different from antibiotics, they don’t drive antibiotic resistance. The problem is that most alternatives like this are more expensive than antibiotics, sometimes by a lot. Individualized treatment wouldn’t solve every problem with a category like bacteriophages, but it would make the economic calculus significantly easier.

Subscribe now

Localization versus Individualization

The mechanics of individualized treatment are yet unsolved, although I’d love to see someone try to tackle it. The first design constraint is flexible capacity. A flock facing an outbreak might need treatment for thousands of birds in a short window; a healthy flock might need treatment for only a few. That likely rules out anything like a centralized robot that goes around giving shots, which would either get overwhelmed in high-demand periods or sit idle the rest of the time.

The best paradigm I can come up with right now is using the existing feed and water infrastructure, which already delivers many poultry drugs at flock level. A system that combined PLF-based diagnostics with hardware that could selectively dispense drugs into a specific feeder or drinker would solve the variable capacity problem.

The practical complication is that birds tend to crowd around feeders and drinkers, so it’s hard to ensure that only the target bird gets the dose. Most drugs aren’t harmful at low doses to a healthy bird, so a pseudo-individualized system that targets the right bird and happens to also dose its neighbors would still be an improvement over the current flock-wide approach.

Thus far, I’ve been speaking about individualized treatment, but actual individualization might neither be feasible nor necessary. There might be upsides to this kind of “localized” treatment rather than “individualized” treatment. Disease outbreaks are often highly physically localized, since bird-to-bird spread is a major transmission vector. If there is a bird in a particular area of the barn with a bacterial infection, nearby birds are also at risk, so it might be prudent to treat those birds prophylactically anyway.

It’s hard to fully appreciate the impact this kind of setup could have, because the current economic equation has been the iron law of poultry health management for basically forever. But once individualized treatment is possible, the prevalence of a condition will stop being part of the economic equation, and all that will matter is the treatment cost per bird. I suspect that once these technological capabilities are in place, it will induce a renaissance of innovation around poultry health, both around prevalent issues that are expensive to treat, and around rare issues that currently aren’t worth treating at all.

The Path Forward

This system would bring poultry to rough parity with cattle and swine, where individualized monitoring and treatment are already standard. PLF systems for larger mammals use RFID tags to track each animal, and farmers can more easily take action on this data because there are so many fewer animals, solving the value proposition problem. Unfortunately, RFID tags are too expensive for chickens, who are worth far less per head, and the bird-per-farmer ratio is too high for hands-on individualized treatment anyway.

Bringing poultry to that parity would require significantly more technological complexity than the cattle and swine versions. There are two components of the system, both of which are massively challenging: accurate individualized diagnosis using hardware-light systems like cameras, and individualized or localized treatment. The former is far beyond the capabilities of even the best current PLF systems, and as far as I know, no one is even beginning to develop the latter.

Part of the reason individualized treatment is a missing piece may be a cold-start problem: neither side of the technology can economically justify itself without the other. But once the loop gets started, the two technologies may become mutually reinforcing. Treatment hardware gives current monitoring systems the value proposition they’ve been missing, which could drive adoption and further improvement. The flood of new data could then unlock new drugs and treatments, on top of the existing ones whose economics are now viable. That further increases the value of better monitoring, and so on.

This framing also helps clarify which current PLF projects are most likely to be useful: the ones moving toward an individualized paradigm. Some current projects use audio monitoring for early disease detection (chickens cough when they’re sick, just like humans), or optical flow algorithms to track lameness at the flock level. These can be valuable, but they don’t move us toward the ultimate prize of a fully individualized system.

I’ve been accused of naiveté for believing that interventions like this could push broiler mortality to sub-one percent. But that’s exactly the trajectory human medicine took. In 1800, roughly half of all children died before reaching adulthood; today that figure is around 4%, and well under 1% in wealthy countries. One of the main drivers was simply being able to figure out what was wrong with a specific patient and treat them for it. Poultry health is still waiting on both. Once diagnosis is a matter of compute (read: free) and treatment is a supply chain problem of getting the right drug to the right bird, there is no physical law that pins poultry health at 6% mortality.

This mismatch leads to a host of challenges for both efficiency and welfare. For one, breeders are significantly less productive than layers: A breeder may lay around 140 eggs over its 60-64 week lifespan, while a layer that’s optimized for egg production can lay up to 400 eggs over its 100 week lifespan.

But the challenges go beyond that. The push towards fast and efficient growth means that broilers eat voraciously. Broilers are slaughtered after only a few weeks, which is too early for any health-related issues deriving from their appetite to matter much for production. But broiler breeders live for much longer, and need to be kept healthy the entire time. Left to eat freely, they would become obese, metabolically dysregulated, lame, and reproductively compromised.

The solution to this historically has been feed restriction, often to around 50–75% of what the birds would voluntarily consume. Restricted feeding keeps the birds alive, productive, and healthy enough to lay the hatching eggs that become the broilers we eat. However, it also means the birds spend most of their lives hungry.

Chronically hungry birds show stress-related behaviors like feather pecking, cannibalism, and frenzied feeding. They often overconsume water to create a feeling of fullness, which worsens litter quality, ammonia, and footpad dermatitis. Stressed birds also lay fewer eggs, which directly impacts production economics.

Until recently, this has seemed like an intractable problem – the incentive to push broilers towards fast, efficient growth is precisely the thing that leads to chronic hunger in breeders. However, more recently, multiple technological solutions have emerged. Layers Laying Broilers, written about previously, uses gene editing techniques so that gene edited layers can lay non-edited broilers.

Another potential solution to make breeders feel less hungry is to just give them Ozempic.

How GLP-1 could work in poultry

GLP-1, or glucagon-like peptide-1, is a hormone found naturally in both humans and chickens. In humans, GLP-1 receptor agonists like semaglutide and tirzepatide have transformed the treatment of obesity and type 2 diabetes. They work through several mechanisms, but the most relevant one here is simply that they make people feel more full.

This mechanism could be useful in poultry as well, but for a slightly different reason. In humans, these drugs help people eat less, but in broiler breeders, the goal would be to make birds feel satisfied with the restricted rations they already receive.

There’s already academic evidence that GLP-1 suppresses appetite in chickens. Researchers, led primarily by Kazuhisa Honda at Kobe University, have demonstrated that administering GLP-1 to broiler chicks reduced food intake by more than 50%, demonstrating that the underlying biology of GLP-1 is conserved across species.

There’s also a particularly interesting hypothesis that breeders may be especially responsive to GLP-1. Since they’re bred to eat constantly, their bodies might produce less GLP-1 naturally, meaning that a small supplement could go a long way.

However, the literature is still young and there’s still a lot that hasn’t been tested. For example, existing studies used the GLP-1 peptide itself, not the longer-lasting GLP-1 receptor agonists like semaglutide that have become blockbuster human drugs. That’s an important distinction, because receptor agonists are engineered to persist in the body for days rather than minutes, which is what would make a practical dosing regimen possible. Additionally, existing studies injected the drug directly into the brain, which would be infeasible at commercial scales.

We also don’t know how mitigating hunger will affect production. In theory, less stressed birds should be more productive, healthier, and less injurious to each other. There’s also a potentially important, but speculative effect where chronic stress in breeders is epigenetically transferred to broilers, making their offspring less healthy. Given that each breeder produces over 100 offspring, even a small improvement in broiler health through a reduction in breeder stress could have a massive compounding economic impact.

Given that the basic idea seems to have validation in the literature, and there are possible economic benefits, the natural next question is how much it might cost.

Costs

“The second pill costs them 4 cents, the first pill costs them $800 million dollars.” – Josh Lyman

Human GLP-1 drugs are famously expensive. Ozempic and Wegovy list for hundreds or thousands of dollars per month in the United States, whereas the total economic value generated by a single broiler breeder is on the order of tens of dollars total.

But the retail price of human drugs doesn’t actually tell you much about the manufacturing cost of the active pharmaceutical ingredient (API). The retail price reflects amortized R&D costs, regulatory expenses, patent-protected margins, marketing spend, and most importantly, the exceptionally high willingness-to-pay of a human patient (or their insurer) trying to manage obesity or diabetes. In a different market like poultry production, where the drug already exists, the economic picture changes.

Recent analyses estimate that the actual manufacturing cost of semaglutide is around $40,000-70,000 per kilogram. That’s already orders of magnitude below what patients pay. And these costs are falling: GLP-1 drugs are the fastest-growing drug category in history, and both Novo Nordisk and Eli Lilly are racing to build out peptide manufacturing at massive scales. Oral formulations with much higher API volumes, and biosimilars expected after patents start to expire in the next few years, will further accelerate that buildout. The cost floor in the future could be dramatically lower than it is today, potentially opening up new applications, including poultry.

Human GLP-1 and chicken GLP-1 share approximately 87% amino acid identity, high enough that one might expect semaglutide or tirzepatide to work in chickens, even though they use the human-specific peptides. If chickens can use exactly the same compounds as humans, they get to free-ride on the enormous economies of scale of human pharmaceutical manufacturing. That’s a massive cost advantage over having to develop and manufacture a chicken-specific peptide from scratch.

Poultry producers obviously aren’t going to pay human drug prices for their birds. But if a drug already exists and is already being manufactured at scale for the human market, selling it into a lower-margin agricultural market is pure upside for the manufacturer — they just need to cover the marginal cost of production. The question is whether that marginal cost will be low enough for a breeder farmer to justify the expense. And that depends on how much production value the drug actually creates, as well as how much it costs to administer.

Share

Getting the drug into the bird

Unlike humans, chickens can’t autonomously take GLP-1s on a regular schedule. The farmer has to administer it to them, which is challenging because there can be thousands of breeders on a given farm. There are really only two options: periodic injections, or mixing the drug into feed or water. Each has different pros and cons, and which one wins could depend on how cheap the API gets — which, paradoxically, depends more on what happens in the human GLP-1 market than anything to do with chickens.

The case for injections is around API efficiency. Oral GLP-1s in humans require roughly 100x more API than injectable formulations, because most of the drug is destroyed in the gut before reaching the bloodstream. We don’t have direct bioavailability data for chickens, but it’s reasonable to expect similar dynamics. If API costs remain high, or if chickens end up needing a chicken-specific peptide rather than a human one, injection may be the only route where the economics work at all.

Periodically injecting so many birds on a regular schedule might sound logistically infeasible at first glance, but there are existing technologies (e.g. this) that are designed to automate the process of on-farm vaccination, which could be repurposed for GLP-1s, Also, keep in mind that broiler breeders are the parents of the chickens we eat, so there are significantly fewer of them – the US breeder flock size is around 60 million, compared to 9 billion broilers produced each year. In reality, the logistical challenge translates to labor costs. In our economic modeling, the dominant cost of administering GLP-1s via periodic injection is the labor needed to operate the automated vaccination machines, not the API.

The case for oral administration is about simplicity. Water-based delivery looks especially promising: a stable water-soluble formulation could be administered directly via the barn’s existing water line, which is already how many medications and additives are delivered on poultry farms. Controlling dosages might be an issue, but breeders tend to over-drink when hungry, so mitigating this hunger could be a natural way to control the amount of water birds drink each day.

Feed-based delivery is less attractive, partly because GLP-1 bioavailability in humans requires fasting (obviously incompatible with taking the drug with food), and partly because many breeders are on skip-a-day feeding schedules, meaning they wouldn’t get the drug on the days when their hunger is highest.

The challenge is that oral methods will use significantly more of the drug. If manufacturing for the human market continues to scale up and costs keep falling as expected, this might be surmountable. But if they don’t, injection might be the only viable path.

There’s still a lot we don’t know about whether this will work, but these gaps point toward a clear research agenda. We need to test whether human GLP-1 agonist drugs have the desired effects in chickens, quantify the production benefits (including possible epigenetic benefits to the offspring, which is where the real money is), and figure out the best administration method. But once all these are in place, it might just be a matter of waiting until manufacturing has scaled up enough in the human market to make chicken use economical. And if this happens, it could help solve many of the stress-related production problems that have plagued breeder farmers for decades, improve the health of billions of broiler offspring, and mitigate a major welfare challenge for their parents.

Innovate Animal Ag is actively exploring ways to accelerate work in this area. If this is exciting to you, or if you have ideas, data, or technical expertise that could help move it forward, please get in touch. And if you want to help build this kind of work full-time, take a look at our open roles.

First housekeeping note - Innovate Animal Ag is hiring again! Our open roles are here. Talent is our #1 determinant of success, so please think about if there’s anyone you know that could be a great fit for any of our roles 🙏.

Second housekeeping note - I turned on paid subscriptions because my understanding is that this is good for the algorithm. All posts here will continue to be free, but if you feel inspired to support our work, all subscription revenue goes directly to IAA, and it could help other readers find our work. In either case, thanks for reading!

Bulwark Biologics, one of Innovate Animal Ag’s flagship projects, started because someone on our team found a paper. It was over a year old, had almost no citations, and described a technology that could meaningfully reduce lameness in broiler chickens. The underlying science (electron-beam inactivated vaccines) had been demonstrated decades earlier in a different application, and USDA had even held a patent on it that they let expire due to lack of commercial interest.

That discovery raised an uncomfortable question: how many other technologies like this are sitting in the literature right now, waiting for someone to notice them?

This question seemed difficult to answer definitively, at least until the entire internet became obsessed with Claude Code a few months ago. We, along with everyone else, started thinking about how to accelerate our work through software projects that previously might have been too difficult or costly. The result was a technology evaluation system that we’ve already used to look at 24,000 patents, and it’s surfacing promising technologies we’d never have found on our own.

Automating Our Own Judgement

When we evaluate a new technology, we ask two questions: 1) Could this have a strong business value proposition for the farmer or producer? And 2) Could this meaningfully improve animal health or welfare? These sound simple, but there’s a lot of nuance behind them, informed by years working within the industry and seeing how many different technologies and businesses are succeeding or failing.

If you asked a current AI chatbot to evaluate a technology against these criteria, it would do a decent job, maybe 80%. But the last 20% is where most of the value is. There are plenty of technologies that look unremarkable on first pass but hold real promise with deeper consideration, and otheres that look exciting but face obstacles only an industry insider would recognize.

To get at this last 20%, we need a process that can actually capture the nuance of our manual evaluation. The problem is that nuance is hard to specify upfront. It’s difficult to write down a sufficiently detailed rubric to hand to the AI. What we can do is show it examples, let it try, tell it where it's wrong, and watch it get closer. That's the core idea: an iteratively calibrated agent that progressively learns to approximate our judgment. Here's how it works:

• We start by writing out our evaluation framework as explicitly as possible and give it to a naive agent as a prompt. The agent scores each technology between 0 and 1 and outputs its reasoning — what arguments it found convincing, what sources it drew on.

• We then develop a set of test cases around technologies we’ve already deeply evaluated where we have strong views about how promising they actually are. For each test case, we define an acceptable range of scores for the agent to output.

• We then run the agent on these same cases, blind to the expected ranges, and compare their output to the desired scores. When the agent’s score diverges from ours, we interrogate its reasoning. Occasionally, the agent weighs a factor we hadn’t considered, and we update the expected range in the test case. More often, the agent misses things, or reasons differently than we would have. For example, it might not immediately understand how poultry integrators make purchasing decisions around issues affecting contract growers, or how to weigh the differences between a technology that merely detects a problem versus one that treats it. In those cases, we critique the reasoning in natural language directly in a Claude Code session, and Claude updates the agent’s prompt for us.

• We repeat this process until all of the test cases pass, meaning that the agent is able to approximate our judgment, at least for a subset of cases.

Once the agent is calibrated, we release it into the world. Our first task was to evaluate every US patent ever filed on an issue relevant to poultry health or welfare — over 24,000 in total. No human team could do a search that comprehensive, but once your judgment is measured in compute costs rather than payroll costs, that kind of depth becomes feasible. The first full run cost a few thousand dollars and took only a few hours.

Critically, every new technology the agent encounters is an opportunity for further calibration. The first patent evaluation run surfaced issues our test cases didn’t cover — for example, how should the agent weigh the existence of competition when evaluating a new technology? When the agent reaches a different conclusion than we would, that becomes a new test case, and we run the calibration process a few more times.

Indeed, the system didn’t work very well out of the box. There was often more nuance that we didn’t think of until the agent explicitly missed it. But after a few rounds of iteration, the system started to work, and it’s started to surface opportunities that we may have never found. Its favorite technology in the existing patent literature is using CoQ10, a cheap compound frequently used in human health supplements, to mitigate ascites in broilers, especially in farms at high altitudes. Multiple patents have been filed with promising data, some over 25 years old. Yet, no one has ever commercialized a product for poultry.

There may be good reasons no product exists. But if there is, it’s likely because of non-public information that the agent wouldn’t have access to, or because of considerations that aren’t immediately obvious to us. Therefore, the process always bottoms out with humans. The purpose of the system is to bubble up opportunities that are worth investing human time into. Then, truly assessing the promisingness of the technology involves having conversations with the people developing the technology and the customers that are actually facing the problem on the ground. If the CoQ10 project ends up panning out, I’ll definitely write about it here.

Our AI Future

The process of iteratively calibrating an agent against pre-defined test cases is one I think applies far beyond this one context. Essentially, it’s a way of automating and approximating complex judgement – like an SOP for cognitive labor. The same architecture could be used for any domain where deep expertise exists, but infinite bandwidth to apply it doesn’t.

For us, the more immediate question is: what else can we do with our calibrated agents? Looking at papers and non-US patents are the obvious next steps, but there could be even more creative applications. One that I find particularly fun is having a “lateral thinking agent” that generates novel business concepts from random word pairings, then runs them through the same calibrated evaluator. Most of the output will be garbage, but if you generate thousands, or tens of thousands of ideas, maybe one of them will be good. Or, you could design another agent that looks at broader technological advancements in biotech or AI, generates possible applications for animal agriculture, and feeds those back through the evaluator.

We’re early in figuring out what these tools can do, and every week we find new ways to use them. We’re looking for an AI engineer to help build out more of these tools, so if this work sounds interesting to you please reach out!

The American response was to recommit to winning the space race, and to double down on investing in innovation. The following year, the Defense Advanced Research Projects Agency (DARPA) was founded with an explicit mission to prevent technological surprise and maintain American superiority. This new agency became a seemingly endless source of breakthrough innovations, contributing to pioneering research in the internet, GPS, modern semiconductor design, renewable energy, and machine learning. Competition, it turns out, breeds excellence.

The outsized productivity of DARPA as a hub of technological innovation is often partially attributed to its unique organizational structure. Rather than operating as a hierarchical bureaucracy like most government entities, DARPA functioned as a relatively loose federation of maverick inventors, innovators, engineers, and entrepreneurs, each granted significant autonomy and resources. The underlying HR thesis was to hire exceptional people and let them cook. By no accident, this philosophy would later become a cornerstone of Silicon Valley’s approach to talent.

DARPA’s success inspired numerous attempts to replicate its model. The US government created several spinoffs focused on domains beyond defense: ARPA-E for energy, IARPA for intelligence, and ARPA-H for health. Other countries followed suit, with the UK establishing ARIA and Europe launching JEDI. More recently, ARPA-inspired structures have caught hold outside of government, with organizations like Convergent Research, and Speculative Technologies.

ARPA is also foundational to Innovate Animal Ag, the organization I founded and currently run. Innovate Animal Ag exists to accelerate breakthrough innovation in animal agriculture, with a focus on animal health and welfare. Given this focus, we sometimes aspirationally refer to ourselves as the DARPA for chickens.

Organizations adopting ARPA-like models vary considerably in their implementation, so when I say Innovate Animal Ag is inspired by DARPA, here are some things I mean specifically:

• We’re organized largely into discrete program verticals, each consisting of small teams with significant autonomy and accountability for success within their domain. These verticals operate mostly independently while still creating opportunities to share strategic lessons, networks, and expertise across the organization.

• We focus on high-risk, high-reward breakthrough innovations rather than incremental improvements. This approach necessarily means some projects will fail, but we view that as a worthwhile cost of a hits-based strategy.

• We’re intervention agnostic. Each innovation faces unique barriers depending on its stage of development and the specific challenges of its domain. Our process begins with rigorously identifying the bottlenecks for a particular technology, whether in R&D, commercial validation, regulatory approval, or industry awareness. After we’ve developed a hypothesis around this bottleneck, we then start to develop interventions to alleviate it. This might mean funding academic research, setting up commercial trials, publishing market research, or facilitating partnerships, depending on what the situation demands.

• We have a strong emphasis on technology adoption, not just research. The most sophisticated technology means nothing if it never reaches the field, so we work hands-on with industry partners to understand concrete needs and accelerate real-world implementation.

ARPA-C

So how has Innovate Animal Ag put this model into practice within animal agriculture? Our first program area was in-ovo sexing, which we began working on in early 2023. At that point, we set an ambitious goal: bring this technology to the United States by the end of 2024.

The first step to making this happen was to deeply understand the current bottleneck of the technology. In-ovo sexing had already been successfully deployed in Europe, placing it in the “early adopter” stage of commercialization, yet it had not spread to other countries.

When we talked to egg companies in the US about the technology, we were surprised by what we heard: basically no one knew that the technology was already succeeding in Europe. This pointed to a clear bottleneck, since obviously no one would adopt the technology if they didn’t know it existed. And fortunately, it was a bottleneck that was relatively easy to alleviate. Over the course of 2023, we published research on the technology rollout in Europe as well as how American consumers were seeing the technology. We helped the poultry industry understand the technology through the trade press and by directly consulting with multiple egg companies in the U.S.

As far as getting companies to adopt the technology, in-ovo sexing was actually an incredibly easy sell, which is to the credit of the early adopters. NestFresh in particular is a brand I want to give a lot of credit to. They already believed that the challenge of chick culling was ethically important to solve, and they know the issue was deeply resonant with their consumers, so as soon as they understood the readiness of the technology they started to mobilize.

We worked closely with NestFresh over the course of 2024, and in December the first in-ovo sexing machine was installed in the US, meeting our initial goal just barely in time. The system became fully operational over the course of 2025, and the first eggs from in-ovo sexed hens reached shelves in July. The technology is now rolled out nationally, meaning American consumers now can purchase eggs from flocks where no male chick culling occurred.

After our initial success spreading in-ovo sexing to the US, we thought that it might be facing a similar bottleneck in other countries around the world. We ran the same playbook in Brazil in late 2023 and in Australia earlier this year. As a result, the first in-ovo sexing machine started operating in Brazil in July of 2025.

As in-ovo sexing gains traction in the United States and Brazil, the technology is transitioning from early adoption to scale-up. This shift means that the bottlenecks are changing as well, requiring us to adapt our strategy. We are currently developing interventions to accelerate scale-up until in-ovo sexing becomes standard practice worldwide.

Within the scope of possible interventions, this kind of market research and consulting perhaps wasn’t the most DARPA-y, but it was in the spirit of doing whatever most effectively unblocked technological progress. It was developing our subsequent program areas that brought us more closely in line with the traditional DARPA model.

The second technology where we invested heavily is electron beam (E-beam) inactivated bacterial vaccines. In 2012, the FDA restricted the use of antibiotics in animal agriculture as a measure to curb antibiotic-resistant bacteria. While this policy succeeded in reducing overall antibiotic use, it also left farmers with one fewer tool to combat bacterial infections in their flocks.

One of the biggest welfare issues for chickens that we use for meat is bacterial infections in their legs, which result from poor leg health due to their fast growth. Hundreds of millions of chickens die each year from a condition called “BCO lameness” before even being turned into meat.

E-beam technology offers a promising solution. Long used in food safety and medical sterilization, it has only recently become economically viable for animal health applications. E-beams kill bacteria by destroying their DNA while preserving surface proteins. These inactivated bacteria can then be injected directly into developing chicken embryos, where the preserved protein structures enable the chicken’s immune system to mount a robust protective response. This represents a significant advantage over traditional chemical inactivation methods, which destroy the bacterial cell structure and produce weaker immune responses.

Academic studies from the University of Arkansas have shown that E-beam inactivated bacterial vaccines can reduce BCO lameness by 50%, and there’s good reason to believe it will be even more effective with refinement. Most promisingly, vaccine doses could be manufactured at scale cheaply enough to have a strong ROI based on reducing pre-slaughter mortality for chicken producers. However, before Innovate Animal Ag got involved, the papers were gathering dust, and the technology remained stuck in the valley of death between proof of concept and commercial deployment.

Unlike in-ovo sexing, which needed awareness and early adopters, electron beam vaccines required commercial validation before industry adoption could occur. Our interventions reflected this different bottleneck. First, we partnered with the University of Arkansas to fund additional research into optimizing vaccine efficacy. Second, we are establishing commercial trials to test vaccine efficacy under real-world conditions with poultry producers, moving the technology one step closer to widespread adoption.

Subscribe now

Looking Ahead

In-ovo sexing and electron beam inactivated bacterial vaccines represent our two largest program areas. Beyond these, we have also incubated a startup currently operating in stealth mode, where we identified a whitespace opportunity, developed the concept, recruited an external CEO, and helped secure initial funding.

We arrived at these focus areas through an extensive process of trial and error. Over the last few years, we explored numerous technologies including things like on-farm hatching, immunocastration for pigs, bird flu vaccination, percussive stunning for fish, and AI welfare monitoring. Some of these efforts failed to gain traction (remember, there will be some flops!). But often the process of failing deepened our understanding of the problem space, and suggested new ideas to trial in the future when we have more resources and capacity.

This learning process has positioned us well for expansion. We have a long list of ideas to try, plus direct experience accelerating innovations at three distinct stages of commercialization: early-stage research and development with our stealth startup, commercial validation with electron beam vaccines, and early adoption through scale-up with in-ovo sexing. With this experience and the DARPA model guiding our growth, expanding our portfolio of program areas will be our main focus in 2026.

As a nonprofit, nothing we accomplished in the past, nor that we’ll do in the future would be possible without the generous support of a large community of people who are involved in many different kinds of ways. If you are as excited about our growth as I am, there are a few different ways that you can support us:

• We’re hiring for a Program Director role. This person will build and run new program verticals, functioning much like a DARPA Program Director. Being a regular reader of The Optimist’s Barn is definitely a positive signal for being a good fit for the role, so I strongly encourage readers here to apply. We’re offering a $2K referral bonus for a warm intro to a candidate we end up hiring, but we otherwise would not have met.

• As we approach giving season, please consider making a contribution to support our work. Right now, the marginal dollar will go to hiring additional program roles, and directly funding technological development. Given how little is currently invested in innovation within animal agriculture, a small amount goes a long way. For example, typical research grants for top universities are often less than $100K, and recent cuts to federal research funding coming from USDA has made funding even more difficult to come by. Therefore, with fairly modest capital deployment, we can become the most important funder of livestock science research in the country, giving us tremendous ability to shape the research priorities of the industry towards addressing the most important welfare issues.
  
  If you’re interested in giving, but want to first learn more about our program and room for more funding, feel free to reach out to me directly. As perennial believers in the power of aligning incentives, if a reader introduces us to a funder that we otherwise wouldn’t have met, we’ll offer a commission to the reader of 1% of the first-year donations from that funder.

• Finally, if you feel like you’ve gotten value from this Substack or its associated Podcast, please consider sharing with others who you think would benefit from it.

In my career, one thing I’ve learned about technological progress is that it doesn’t happen by default, especially in a sector like animal agriculture. Progress requires a community of dedicated evangelists and innovators who are obsessed with pushing forward the frontier. So thanks for being part of our community!

Unfortunately, the reality of the situation is not as triumphant as it seems. In fact, we’re on track to continue to make the same mistakes we’ve been making for the last three and a half years. Bird flu is now endemic in wild birds, meaning it persists year-round in wildlife populations, and it’s cyclical, meaning that it’s more extreme when migratory birds are in your area.

Every fall since the disease re-emerged in 2022, bird flu cases have spiked due to migratory patterns of wild birds. Each time this happens, the population of egg laying hens contracts, egg production decreases sharply, and prices skyrocket. In both 2025 and 2023 prices got high enough to make mainstream headlines. Innovate Animal Ag estimated that bird flu cost American egg consumers $14.5 billion in higher prices from June 2024 - June 2025.

Now, as we move into another fall season, bird flu cases are again starting to increase, suggesting we may be headed into another ramp-up in infections. If rising egg prices catch us off guard this winter, it’ll be quite an achievement—after all, we changed nothing and somehow expected a different result.

Just in the last few weeks, 2 million layers had to be depopulated in Washington, and 3 million had to be depopulated in Wisconsin. For context, the total size of the US layer flock is 300 million, so just those two infections alone are responsible for decreasing the US’s egg production capacity by 1-2%. When bird flu suddenly contracts the size of the national layer flock, that’s what causes sharp increases in the price of eggs.

That said, I can’t say for sure that egg prices will definitely spike again in the coming months. Egg producers have been proactive about repopulating the US layer flock, and in particular have been increasing the number of layer breeders (the parents of the hens that lay the eggs that we eat) that can quickly create more layers when needed. Also, in the last year companies have invested heavily into improving their biosecurity, which could help prevent infections. This all could make 2026 more like 2024, when egg prices increased a little bit, but not enough to make major headlines.

However, biosecurity will never be a full solution, especially since the CDC confirmed that HPAI is now capable of airborne transmission (something many poultry farmers suspected long before the official confirmation). Ultimately, if viral particles can drift in through air vents, there’s not a lot that farmers can do.

As I’ve argued before, the one and only long-term solution to bird flu is to allow farmers, particularly in the egg industry, to vaccinate against the disease. Other countries already do this (most recently, South Africa), and in fact we manufacture an effective bird flu vaccine right here in the USA that we currently export to other countries.

Unfortunately, farmers are only allowed to administer vaccines that USDA approves, and USDA has not yet approved a bird flu vaccine. The rationale is complicated, having to do with the international trade of animal products. The concern is that vaccines might suppress symptoms but not eliminate the virus entirely, so if you import from a country that vaccinates, you risk accidentally bringing in the virus into your country. Because the US is a major exporter of chicken meat, we don’t currently allow any poultry producers to vaccinate.

There are clear policy changes we can make to improve the situation. Bird flu primarily affects egg laying hens, and not chickens we use for meat. However, we mainly export chicken meat, not eggs. These two supply chains are completely distinct, and in theory we should be able to vaccinate egg laying hens without it affecting meat exports at all. In practice, trade agreements are complicated and it doesn’t yet work this way. But if the Trump administration is serious about keeping egg prices down, they should make this a priority in current ongoing negotiations with our trade partners.

When egg prices were in the news earlier this year, it coincided with a change in administration. Trump mostly blamed Biden’s failed bird flu mitigation policies for the problem. At that point, Trump’s criticism was justified, and I said so. Unfortunately, Trump’s policy has been largely the same as Biden’s, which will have predictably similar results.

To the administration’s credit, there have been some positive rumblings on potentially changing vaccination policy, but nothing yet concrete. And given the logistics of how a vaccine would be administered, it’s likely too late to affect where egg prices will go this winter.

If egg prices do start making headlines again, we should resist the urge to lump the story in with other concerns swirling around right now about cost of living. Bird flu is a unique situation that has to do with veterinary and trade policy, not macroeconomics. I can only hope that the Trump administration is able to make this distinction, and implement a rational, science-backed policy response. Allowing vaccination would be a rare policy win that benefits consumers, farmers, and animals alike.

My Favorite Post

• Don’t Demonize Efficiency in Animal Agriculture

People often say that the best reason to write is to figure out what you think.

I didn’t find this to be true when I started The Optimist’s Barn. The first chunk of essays I wrote were mostly things that were already in my head, and it was just a matter of writing them down.

But Don’t Demonize Efficiency in Animal Agriculture was the first time I feel like I really convinced myself of something through writing. Before I read Mike Grunwald’s book We Are Eating the Earth, my take on the relationship between welfare and efficiency was that it’s complicated. The two are sometimes aligned, and sometimes not, and it’s hard to say which one without looking at the specifics. An earlier review I wrote of a different article by Grunwald articulated this perspective.

But after internalizing the arguments in Grunwald’s book, and deeply thinking through the analogies to animal welfare, I’ve moved closer to the view that on the current margin they’re usually in alignment. If we take current chicken demand as constant, and compare today’s agriculture to 60 years ago, improvements in efficiency mean that today we need 3.3 billion fewer chickens and half as many chicken-days to meet current demand.

When I then wrote about how livestock breeders have saved more animal lives than animal rights advocacy ever has, I was of course being a little provocative, and obviously keeping demand constant isn’t the most robust way of doing economic modelling. But I feel like there is actually something deep and important about this that lots of people get wrong.

My current mental model is now something closer to the following: there were ways in which farming was higher welfare in the olden days, when a huge percentage of the population was engaged in farming, and farms were relatively small. But this system was never going to survive the society-wide transition to an industrialized economy. We just need too much food.

The move from pre-industrial to industrial farming was a step change, and in that change lots of stuff went wrong. For example, in 1925, 18% of chickens died before making it to slaughter, which says something about the kinds of conditions they were living in. But in the subsequent decades, farmers gradually learned how to keep animals healthier and productive, since it was in their economic interest to do so, and mortality fell to a low of 3.7% in 2013.

And after industrialization is baked in, the economic incentives of farmers started to re-align with animal welfare. The rationale is simple: if production causes X amount of negative externalities, and we double efficiency, then we need half as much production, yielding X / 2 externalities. This isn’t literally true in every case, and stocking density, in particular, might be a notable exception. But if Grunwald is right and one of the best ways to preserve the environment is to farm as intensively as possible to use as little land as possible, a similar argument might hold for animals as well. Likely, many farmers would think this is obviously the correct way of thinking about things, and would be frustrated that it would seem so contrarian in some circles.

This model also provides an answer to a question I often get, which is “if X technology is economically beneficial, why hasn’t it already happened?” The answer is that agriculture is in a continual process of improvement, and there are always new innovations that make things better on the current margin.

All this isn’t radically different from what I thought before, but I do feel like it’s an important deepening of my understanding of the world. And to writing’s credit, only by sitting down and forcing myself to do a book review did this new model become crystalized.

My Most Underrated Posts

• Why the New Administration Should Consider Farm Animal Welfare

• Why PETA’s Campaign Against Certifiers is Misguided

Looking back, I also feel particularly proud of both these pieces. I think they each point to something important, true and underrated, and do so in a persuasive and clear way. But neither performed very well in terms of views or likes. The one about government policy is my second least read post.

I suspect that one of the reasons that both underperformed has to do with the titles. Early on, I wasn’t sure how to create appealing titles (which, for better or worse, is probably one of the top determinants of a post’s success). Some titles, like the government one, were more descriptive, some, like the PETA one, were more explicitly clickbaity.

I thought that leveraging controversy around PETA might be a good way to hook into a piece. People on the internet love controversy, right? But the essay fundamentally isn’t about PETA at all, but rather about the role that certifiers play in the meat supply chain. In this case, conventional wisdom about how to get eyeballs on the internet led me astray.

I think the reason is that, truthfully, you all are too smart.

I first started to realize this when I published the piece Individuation is a Key Challenge in Animal Welfare Technology. To be honest, I didn’t feel great about this post before I published it. I even told a friend the night before that I thought it was my weakest piece so far. Then, it was my first piece to hit over 1K views.

I was initially very surprised by this. I don’t think “individuation” is even a real word, I’m pretty sure I made it up!

But then I realized that maybe I was underestimating how my audience differs from the broader internet. At the risk of overflattering you, I don’t think you come here for oversimplified explainers and easy-reading listicles. I think you come here for nerdy, technical deep dives into complicated and nuanced issues. And the word “individuation” is nerdy, technical, and nuanced. It’s cool to write for such a sophisticated audience!

Therefore, I’m hereby renaming the pieces, in case anyone wants to give them another look:

• Why the New Administration Should Consider Farm Animal Welfare is renamed to “Explaining the Vote-Buy Gap in Animal Welfare: Why Subsidies Make Better Policy.” It argues that modeling animal welfare as a collective action problem rather than an individual action problem explains the vote-buy gap, and why under this model animal welfare subsidies can be more economically efficient and politically expedient than bans.

• Why PETA’s Campaign Against Certifiers is Misguided is renamed to “The Animal Welfare Premium is not Captured by Certifiers, but AI Can Help.” It argues that there is clear market evidence that there’s money to be made in higher-welfare products, but that this premium is captured by brands, not certifiers. It then argues that if certifiers operated more like for-profit businesses, potentially aided by AI, that could lead to a more efficient market that’s better for farmers, certifiers, and animals.

Growth

This is especially navel-gazy so feel free to skip, but I know a lot of folks here are interested in building audiences of their own, so I thought it might be helpful to share.

Over the course of the last year, the Substack has grown from 0 to 810 subscribers. Substack also provides a “Total followers” metric, which I don’t really understand, but which is currently around 3,000. By far the biggest growth channels were recommendations from other Substacks, mainly Astral Codex Ten and Asimov Press (thanks!).

Growth has been fairly steady, but with a few inflection points. The first inflection point after the initial growth spurt came from two semi-high profile reposts: one from Mike Grunwald for Contra Grunwald on Industrial Farming, and one from Richard Hanania for Animals Need Us to Be Rich. The second big inflection point came, surprisingly, after I started recommending some other Substack’s from The Optimist’s Barn. I’m guessing this had something to do with teaching the Substack algorithm who my audience was, but I’m not totally sure. I’m also not sure why it levelled off so suddenly after ~3 weeks.

Looking Ahead

As most of you know, my main job is running Innovate Animal Ag, a think tank that puts into practice many of the ideas I discuss here. The reason I started the Substack is because I felt like there were a lot of assumptions and ideas behind our theory of change that didn’t have a natural forum. I tried to express them through X, or in one off pieces like the one I wrote for Asimov Press. But I never really felt like there was enough time and space to say everything I wanted. Then, I read Go Direct: The Manifesto by Lulu Cheng Meservey, and decided I needed to build my own audience. Substack seemed like the most natural fit.

After a year of publishing at roughly a cadence of 2 posts per month, I feel like the Substack is at a bit of an inflection point. Growth has plateaued over the last few months, and I’m starting to run into the classic problem of finding it harder to find topics to write about. I’m not totally tapped out of course, but I feel like I have two options. One is to double down, continue publishing as much as possible to get out of this growth plateau, and potentially start a paid tier that would support Innovate Animal Ag. The other is to stay at roughly our current reach and decrease the amount of time I spend writing.

Things at Innovate Animal Ag are going extremely well, and I feel like shepherding the organization through an incoming period of expansion is going to require as much of my time as attention as possible (I’ll write more about this soon!). Therefore, I’m going to shift my policy towards writing only when I have something important to say, rather than trying to publish at a regular cadence.

Derek Thompson once said you don’t really write for your audience, you write for the audience of your audience. And I’m honestly humbled and flattered by the caliber of some of the folks that I know are already regular readers. Seriously, thank you for your time and attention!

One of the companies that bucks this trend is NextHen, a startup whose vision is to rethink how we do poultry breeding. Many of the challenges we’ve discussed on The Optimist’s Barn trace back to genetics. NextHen’s mission is to find practical ways to leverage better genetics to solve some of the biggest challenges the poultry industry faces around welfare, sustainability, and productivity. I sat down with Dr. Yuval Cinnamon, the CSO of NextHen and a Principal Investigator at the Volcani Institute, for a conversation about one of NextHen’s most exciting projects, called Layers Laying Broilers (LLB).

Growth vs. Reproduction

“Biology has a budget,” said Dr. Cinnamon during our conversation. “A human cannot both be a sprinter and also a sumo wrestler.” In other words, a biological organism that’s heavily optimized in one direction will start to perform worse in other areas.

This is the situation for modern poultry. A century of selective breeding has pushed poultry into two specialized breeds: layers, optimized for egg production, and broilers, optimized for meat production. “The performance of the resulting breeds is clearly excellent,” he said, “but it comes with a price. And the price is that the fast growing broilers have very poor reproduction, and layers are excellent at laying eggs but they produce very little meat.” By his estimate, layers are 2.5 to 3 times better at producing eggs than broilers.

The challenge is that we cannot yet fully separate labor in the poultry industry into laying and growing–there still need to be chickens responsible for laying the broilers that are used for meat. These chickens, called broiler breeders, necessarily share genetics with the broilers. Broiler breeders are used for their reproduction, but have genetics optimized for fast growth.

This mismatch leads to a host of challenges for productivity, sustainability and welfare. “If broiler breeders had unlimited feed,” Dr. Cinnamon said, “they would grow so fast and become so heavy that they would not actually reach the age of sexual maturity. Imagine an average baby born at the weight of three kilos and growing to 200 kilos after two months. The metabolism of broiler breeders cannot adapt to this fast growth for too long. Their cardiovascular system will not support it. Their skeletal system will not support it. They will crash. And in order to prevent this, breeders must keep these parent stock under feed restriction from very early stages of development. And this is a major issue because for their entire life, they will be hungry. They were selected for many years to eat all the time, so their appetite is genetically endless.”

Hunger isn’t the only problem. Feed restriction often leads to elevated stress and aggression. Stress hormones in broiler breeds are measurably higher than in their layer counterparts, which are not kept under feed restriction. These hormones may also have maternal effects, meaning they can be transmitted to the offspring, causing the broilers themselves to be more stressed.

These welfare issues go hand-in-hand with lower productivity. A broiler breeder may lay around 140 eggs over its 60-64 week lifespan, while a layer can lay up to 400 eggs over its 100 week lifespan.

Until recently, these challenges with broiler breeders were a necessary part of poultry production, given the unbreakable genetic link between broilers and broiler breeders. But NextHen is seeking to change this with LLB.

Subscribe now

Breaking the Genetic Link

At a high level, LLB uses gene-editing to create layers whose offspring can have distinct genetics from their own. These layers, called sterile surrogates, can then be used as the parents for fast growth broilers. In breaking the genetic link between broilers and broiler breeders, the LLB approach has the potential to completely solve all of the productivity and welfare issues associated with broiler breeders. It combines the best of a layer’s reproductive ability, and the broiler’s growth ability, fully completing the division of labor between the two breeds.

The process starts by isolating primordial germ cells (PGCs) from an existing fast growth broiler line. “The germ cells are precursor cells which give rise to the gametes, the sperm and eggs, and their only role in our body is to transfer the genetic material to the next generation,” explained Dr. Cinnamon. These germ cells are then cultured in-vitro and banked to ensure a stable supply.

Then, NextHen builds the surrogate layer line using a proprietary gene edit that confers inducible sterility, allowing breeding companies to control which birds become surrogates and which remain fertile to propagate the surrogate population.

When the surrogate layers are developing as embryos inside their eggs, the PGCs of the fast growth broiler line are transplanted into the eggs via an in-ovo process, like a more precise version of in-ovo vaccination. As the embryo develops, it treats the transplanted PGCs as it would its own native PGCs.

Once the eggs hatch and the surrogates grow up, they can be raised exactly like any other layer, except their offspring will be fast-growth broilers. These broilers can then be raised in exactly the same way that current broilers are. Although the sterile surrogate line is technically genetically modified, their offspring, the actual broilers, don’t share any of their surrogate parents’ genetics, meaning that the resulting meat can still be considered non-GMO.

NextHen has already successfully demonstrated this concept in a lab setting, and is in the process of developing the technology for commercial application.

The Future of Poultry

LLB is just one of the many projects in NextHen’s pipeline. For example, they also have an in-ovo sexing solution that can remove the need for male chick sorting and culling using similar gene editing techniques to LLB. Through clever breeding strategies, NextHen created a breed of layers named Golda with a gene edit only present on the male sex chromosomes. The gene edit allows hatcheries to selectively halt the development of male embryos (e.g. through exposure to blue light), so that only females hatch. Since the females lack the male sex chromosome which contains the gene edit, they can be treated as non-GMO.

NextHen is also exploring the potential of featherless birds. Feathers trap heat, driving up cooling and ventilation costs, and causing major welfare issues related to heat stress. Additionally, slaughter plants use significant water and energy for plucking. A viable featherless line could reduce heat stress while also lowering energy and water use.

It’s rare to see a company tackle the challenges of the poultry industry in such an innovative way. When I asked Dr. Cinnamon why he’s chosen to tackle these problems, he pointed to the long term importance of the chicken industry to the global food system. Given the superior feed efficiency and sustainability of chicken, Dr. Cinnamon anticipates the industry will continue to grow, especially in markets like India and Africa. “For the foreseeable future, I don’t see a replacement protein source that will replace animals. Animals have been used for food for thousands of years, and it’s a cultural thing. It’s not going to change very fast,” he said. “And until then, we need to raise these animals with compassion.”

My reaction to reading these statistics was that it's a little hard to believe the world actually works this way! The miracle of modern agriculture has given us food so abundant that we go through the trouble of clearing out land, planting seeds, fertilizing, watering, harvesting, packaging, and shipping all this food, only to throw almost a third of it away?

There may be a sense in which food waste is economically “rational,” given that people want fresh food, and food expenditures are a shrinking percentage of the American budget, meaning that the cost of waste is going down. However, the lesson to me is that “efficiency,” in the sense of doing more with less, isn’t always the thing that market forces narrowly optimize for (the chicken industry itself also demonstrates this). Given the myriad negative externalities associated with agriculture, we can and should take action to make it more efficient in this sense.

Fresh meat accounts for only 3% of total food waste, equivalent to 1.84 million tons. But there’s also a hidden multiplier in this figure since animals also need to eat. In fact, most of agriculture in the US actually goes to feed livestock, not humans. When meat is wasted, the corn, soy, and other ingredients that the animal ate during its life is also wasted. I call this “feed waste,” as distinct from “food waste.”

Though unaccounted for in the above statistics, feed waste starts before meat even reaches the human food supply chain. A substantial percentage of livestock animals get sick and die before being slaughtered, which is not only an animal welfare issue, it’s also a waste of all the corn and soy they consumed up to that point. By my calculations, almost 4.4 million tons of feed are wasted in the US each year due to premature livestock death, which is on top of the 72 million tons of human food waste.

It’s useful to zoom in on this part of the supply chain, because the set of possible interventions is different. Averting pre-slaughter feed waste is the realm of animal health, an area particularly amenable to technological solutions. By giving farmers better tools to keep animals healthier, we can reduce feed waste and improve animal welfare in ways that don’t rely on consumer behavior change, restructuring complex food supply chains, or inefficient changes in farming practices like regenerative agriculture.

Breaking Down the Numbers

Each sector within animal agriculture contributes different amounts to this 4.4 million wasted tons of animal feed. The chicken and pork industries are the major contributors, with the beef and turkey industry having smaller contributions.1 At a high level, the impact of each sector depends on the average mortality across the sector, as well as how much feed each animal eats. However, there’s some sector specific nuance that’s worth going over:

• Chicken - Mortality among broiler chickens currently stands at roughly 6%. Broilers are the most efficient animal at converting feed into meat, so each individual chicken consumes a relatively small amount of feed. However, the scale is enormous since Americans eat more chicken than any other kind of meat. Over 500 million chickens die before slaughter in the US each year (which, for perspective, is more than all the other land animals that we actually eat), cumulatively consuming 2.2 million tons of wasted feed

• Pork - The pork industry has the highest mortality rate in animal agriculture, with almost a quarter of pigs dying before slaughter. However, many of these deaths occur early in life during the “wean” phase when they mainly consume their mother’s milk. This analysis only considers post-wean mortality, which sits around 10%, accounting for 1.9 million tons of wasted feed annually.

• Turkey - Americans eat less turkey than other kinds of meat, but they’re worth including here because turkeys are larger than chickens and consume more feed, and because mortality in the turkey industry is relatively high, especially among males. Turkeys are separated by sex, with smaller females often sold as whole birds, and larger males being used for other purposes like cutlets and deli meats. Female turkey mortality sits around 5-6%, while males are at 10-12%, which combines for 386 thousand tons of feed waste per year.

• Beef - Beef cattle have relatively low mortality, at around 3% post-wean. Additionally, beef cattle generally live longer than a year, meaning that annualized mortality is even lower. Part of this is because a single beef cow can be worth upwards of $2,000, creating a strong financial incentive for cattle ranchers to maintain good health. A beef cow might consume 3 tons of feed before slaughter, but low mortality, a long life, and a relatively small number of them means the total feed waste in the beef industry is only 383 thousand tons per year.

Taken together, pre-slaughter death in the livestock sector accounts for 4.4 million tons of wasted feed per year. Using standard estimates of the carbon footprint of corn feed (0.42 lb CO₂e/lb) and soy feed (0.55 lb CO₂e/lb), this translates to 2.4 million tons of carbon emissions for which there was no economic benefit. While smaller than other sources of climate impact from animal agriculture, such as enteric methane emissions or deforestation for cattle grazing land, it’s still a meaningful contribution: roughly equivalent to 500 thousand cars driven for a year. Additionally, these “unproductive emissions” are by definition unnecessary, and with some technological innovation it may be possible to significantly reduce them in market-friendly ways.

Make Animals Healthy Again

One concerning trend is that pre-slaughter mortality seems to be increasing across the meat sector. Broiler mortality has increased from 3.7% to 6% over the last 12 years, mirroring other declines in efficiency across the broiler sector. Post-wean swine mortality has increased from 8.55% in 2016 to 10.49% last year. Similar trends can be seen in the turkey and cattle sectors. Taken together, these trends account for hundreds of millions of animals dying without generating any economic value. The underlying reason for this phenomenon is unclear, but it poses the question: How can we reverse this trend, and ideally make livestock mortality as low as possible?

One of the big challenges with finding interventions to improve livestock health is that it’s already in the economic interest of farmers to keep animals healthy. If an animal becomes sick and dies before its economic value can be accrued, not only is that bad for the animal, it also means that the farmer has lost revenue. Similar to food waste, there is a sense in which this is economically rational: current mortality rates reflect an equilibrium that balances the marginal cost of improving animal health and the economic cost of mortality.

This is why technology is such an important tool for animal health: it can change this equilibrium. By making it cheaper to prevent disease, we can reduce mortality and align the incentives of farmers with sustainability and animal welfare.

One promising example is a novel way to create bacterial vaccines using electron-beams. Bacterial infections are a major cause of livestock deaths, especially in the high-density conditions of modern farms. In recent years, farmers have been under pressure to reduce antibiotic use to slow the rise of antibiotic resistance, which limits their options for fighting these infections.

Electron-beam technology, long used in food safety and medical sterilization, has only recently become affordable for use in animal health. Electron-beams kill bacteria by destroying their DNA but preserving their surface proteins, allowing the animal’s immune system to recognize and mount a more robust protective response. This is a significant advantage over traditional methods of inactivating bacteria with chemicals, which can damage the cell surface, resulting in a less robust immune response.

One potential application of this technology in the chicken industry is to fight a condition called “BCO lameness.” Chickens raised for meat are bred to grow extremely quickly, which puts intense pressure on their legs, often leading to bacterial infection in their developing bones and joints. One major consequence is lameness—poor leg health that makes it agonizing for birds to walk. The pain is often so intense that a chicken will die of thirst rather than walk, so birds with severe lameness must be euthanized once identified. Up to 225 million chickens in the US die each year from this condition, wasting 909 thousand tons of chicken feed.

In preliminary academic studies, vaccines using electron-beam inactivated bacteria have reduced BCO lameness by 50%, and there’s good reason to believe that with further development, the effect could be even larger. Electron-beam vaccines have yet to make it onto the market, but if and when they do, they could be a powerful tool to improve livestock health across animal agriculture. (There are some exciting opportunities right now to advance this technology through either investment or philanthropy, please DM me if you’d like to learn more!)

Technologies like electron-beam vaccines hold tremendous potential to lower the environmental footprint of animal agriculture, while simultaneously improving animal welfare, and boosting farm profitability. Among the many difficult trade-offs in reducing agriculture’s externalities, this is one area where incentives align: no one benefits from animals dying prematurely. The challenge is to make better health outcomes affordable, which is exactly what new technologies can do.

However, when introduced to in-ovo sexing technology, an alternative which allows producers to identify and remove male eggs so that only females hatch, consumer interest is overwhelming. 73% of Americans describe themselves as “extremely” or “very” interested in eggs produced using this more ethical method.

This question is now no longer a hypothetical: consumers now have the opportunity to vote with their wallets to support this new practice. NestFresh is debuting eggs from in-ovo sexed hens under the "Humanely Hatched" label, now available at select Whole Foods locations in the Southwest and soon expanding nationwide. Another brand, Kipster, will follow later this fall with its own line. Keep an eye out for these brands next time you shop—your purchase can help accelerate a critical shift toward more humane egg production.

These launches mark an important turning point for the US egg industry. After many years of scaling up in Europe, where in-ovo sexing is now 28% of the market, the technology has finally arrived in the US. The first in-ovo sexed chicks were hatched last December, and their eggs are just now hitting shelves. NestFresh and Kipster are the pioneers of this technology, and NestFresh in particular deserves credit for actively driving progress in the US. As early adopters, they are taking on some business risk since in-ovo sexed chicks are currently more expensive. This means that these companies had to invest significantly more in their flocks, and their bet is that they’ll be able to sell more of their eggs to recoup their investment. In other words, they’re hoping that the self-reported survey data about consumer interest in in-ovo sexing holds in practice.

Hopefully, many other brands will quickly follow the lead of these two companies. However, it takes six months from hatching for chicks to reach laying age, meaning that these might be the only two brands available in the US for a while.

How male chick culling started, and how it will stop

Like many welfare challenges, male chick culling was a side-effect of industrialization in poultry farming. Historically, chickens served dual purposes: first providing eggs and then eventually being slaughtered for meat. However, in the push towards economic efficiency, the poultry industry split into two distinct markets, one for eggs and one for meat, each using intensive genetic selection to optimize for their distinct economic goals. In egg production, efficiency meant selecting hens that laid as frequently and consistently as possible. In meat production, it meant selecting chickens that grew rapidly, converted feed efficiently into muscle, and produced larger, more desirable cuts like the breast.

This specialization is why chicken meat and eggs are some of the cheapest and most abundant sources of animal protein we have today. But another consequence was that male chickens in the egg industry stopped serving any economic purpose. Since these males neither lay eggs nor efficiently produce meat, hatcheries found it economically necessary to cull them shortly after hatching.

Subscribe now

Clearly, there’s some amount of waste inherent in doing things this way—we incubate and hatch hundreds of millions of eggs only to discard half of them. But this practice persisted because the economic benefits of using genetically specialized breeds outweighed the cost of that waste. The power of technology in this case is to allow us to optimize in both directions at once. In-ovo sexing lets us preserve the efficiency of genetic specialization while eliminating the need for chick culling, offering both ethical progress and, eventually, economic efficiency.

There are multiple approaches to in-ovo sexing. One of the most prominent approaches uses advanced imaging to detect subtle optical differences between male and female embryos. NestFresh’s Humanely Hatched eggs use a technology of this type called Cheggy, developed by Agri-Advanced Technologies, which uses hyperspectral imaging to detect subtle differences in feather color between male and female embryos.

Another widely adopted technique involves taking a small fluid sample from inside the egg to test for sex-specific hormones or genetic markers. Kipster’s eggs, launching later this year, will use a technology of this class employing PCR analysis to detect the sex chromosome, developed by the company Respeggt.

The path to common practice

In the long term, in-ovo sexing is likely to become economically advantageous compared to traditional manual sexing methods, and scale is the key to making this happen. As in-ovo sexing becomes more widely available, economies of scale and technological refinements will help bring its cost down. Fundamentally, in-ovo sexing is an automation technology, replacing highly skilled human laborers with precise and optimizable machines.

There are also co-benefits working in in-ovo sexing’s favor: removing male eggs early frees up incubator space for other productive uses. Additionally, once male eggs can be removed, other beneficial practices such as on-farm hatching and in-ovo vaccination could yield further productivity increases (as well as additional welfare benefits) that help defray the cost of in-ovo sexing.

However, we’re not there yet. In-ovo sexing is still at the beginning stages of its rollout, and while it has become substantially cheaper since it became available 7 years ago, it still adds a few cent per dozen to the production cost of eggs. This is why initial adoption has been limited to premium brands like NestFresh and Kipster, which supply higher-value categories such as free-range and pasture-raised. Customers of these brands have already demonstrated their willingness to pay a substantial premium for eggs produced in a better way.

One of the more underrated challenges of technological development in the physical world is that it’s not sufficient for a technology to be economically advantageous in the abstract. It also needs a path to scale. If we suppose that widespread adoption would, in theory, create the economies of scale needed to make the technology cost-effective, this doesn’t guarantee this adoption happens. There needs to be a clear economic rationale for the first hatchery to install a machine, then the second, and so forth. Each level of scale requires its own economic calculus.

In the case of in-ovo sexing, this pathway to scale exists because there are consumers willing to pay slightly more for ethically-produced eggs. The egg market already has a clearly defined premium structure ranging from conventional to cage-free, free-range, and pasture-raised, with each category reflecting increased prices and higher welfare standards. A new technology like in-ovo sexing can start in the most premium parts of the market, then move downmarket as adoption expands, the technology matures, and costs decline.1

This is why it’s so important that the initial entrants into this new egg category succeed. Forward-thinking companies like NestFresh and Kipster are betting that consumers will reward their investments by buying more of their eggs. But not every company in the egg industry is prepared to take such risks. Many are waiting until there’s clear market evidence before following suit. If the first wave of Humanely Hatched eggs succeeds commercially, other producers will jump in to capture a piece of this premium market, broadening adoption and helping bring costs down.

So next time you find yourself in the egg section, consider purchasing eggs from NestFresh or Kipster. You'll not only have the comfort of knowing that no male chicks were harmed to produce your eggs, you'll also be sending a powerful market signal: ethical egg production matters to consumers. Your purchase will help in-ovo sexing technology scale and make it more likely that it eventually becomes standard practice across the entire industry.

How can this be? Is it because rates of lactose intolerance are rising, or because people started to realize that dairy is a land-intensive process that’s bad for the climate? Or is it a triumph of vegan advocacy that convinced people to consume less milk, and instead drink alternatives like soy, oat, and almond milk?

None of the above–it’s because the average output of a dairy cow increased from 4,572 lbs per cow to 24,178 lbs per cow. That increase came from decades of selective breeding, improved nutrition, better veterinary care, and more optimized farm management. When each cow produces five times as much milk, you need significantly fewer of them

Getting more with less is a central theme in We Are Eating the Earth, where Grunwald offers a sober, realistic, and insightful take on how we can feed a growing population in a world that’s heating up faster than we’d like. He zeroes in on how we tally the climate costs of agriculture, and how bad climate accounting has led us toward flashy but counterproductive solutions. When we do the accounting right, two truths emerge: first, there are no silver bullets in agriculture. Second, increasing productivity is one of our most powerful tools, because it lets us produce more food with less impact.

Focusing on the virtues of agricultural productivity may seem surprising coming from a staunch environmentalist like Grunwald. But it reflects a broader truth: industrial agriculture may cause emissions, pollution, and poor animal welfare, but the more efficient agriculture is, the less of it we need.

The Opportunity Cost of Land

When it comes to climate accounting, Grunwald argues that land use is often the most overlooked aspect. The reason conserving land is important isn’t just to preserve scenic vistas or to protect endangered species. It’s because forests, grasslands, and other natural ecosystems are the best ways we have to keep carbon in the ground rather than in the atmosphere. Converting land to other uses, even purportedly climate-friendly ones, always comes with opportunity costs that many carbon accounting models ignore.

Grunwald goes through much discussed climate solutions like ethanol, biodiesel, and biomass energy, and explains why each is either useless, or even potentially harmful if it requires land to be repurposed. For example, growing corn as a biofuel may be positive on a narrow accounting (because corn takes carbon only out of the air and top soil), but it’s actively harmful if you take into account the opportunity cost of the land on which the corn is grown.

Of course, we need to use some land to feed people. But the less we use, the better. And one of the best ways we have to use less land is to get as much as possible out of the land we do use. This is why solutions like regenerative agriculture also fail as climate solutions. They’re less efficient, require more inputs, and demand more acreage to produce the same amount of food.

To critics of industrialized agriculture, this analysis may be frustrating. They might argue that a ruthless drive towards efficiency is exactly what caused all these problems in the first place. But that perspective misses something crucial: it’s only because agriculture is so efficient that we can feed 8 billion people today without already having cut down the rest of the world’s forests.

To be sure, industrialized farms have a host of problems that smaller farms don’t, both for the climate and for animal welfare. But many of these issues aren’t intrinsic to efficiency itself; they’re unintended side effects of how efficiency has been pursued. The ruthless efficiency of industrial agriculture makes them hard to fix, but in theory, if we can solve them without sacrificing productivity, we can actually make things better.

Efficiency Saves Lives

If flawed accounting can mislead us about climate solutions, it’s worth asking whether we’ve also misunderstood the effects of industrialization on animal welfare. It’s easy to assume that efficiency always comes at the animals’ expense. But the data tells a more complicated story.

One key dynamic is that the more food each animal produces, the fewer animals we need to raise. We consumed roughly 498 pounds of beef per cow in 1960 versus 795 pounds today, 141 pounds of pork per pig in 1960 versus 173 pounds today, and 3.04 pounds of chicken meat per chicken in 1960 versus 4.08 pounds today.1

Of course, overall meat consumption and population levels have both risen, so outside of the dairy sector, the total number of animals hasn’t gone down. But it hasn’t gone up as much as you’d expect. To try to understand how much impact efficiency has had, we can try a simple counterfactual: What if we held today’s demand constant, but produced it using the efficiency levels of 1960?

Under this analysis, we see that increases in agricultural efficiency since 1960 mean that we need 20-40% fewer animals to meet current demand, depending on the species.2 Just in 2023, these efficiency gains would have accounted for 26 million fewer cows, 29 million fewer pigs, and 3.3 billion fewer chickens than would have been needed with the agriculture of 1960. On this method of accounting, efficiency gains have averted more animal deaths than any sort of climate or animal advocacy ever has.

Of course, that isn't the only way to do the welfare accounting. For one, efficiency also lowers costs, which can drive up demand. And comparing agriculture now to agriculture in the 1960s isn’t apples-to-apples. There are many health and welfare issues that animals face now that they didn’t in the past. For example, the increased size of the average chicken has led to painful heart and leg problems that were far less common in historical breeds.

At the same time, we also can’t necessarily claim that welfare is worse in every way now than it was in the past. One meaningful metric is “mortality”– the number of animals that die before reaching slaughter. In 1915, mortality for broilers was a whopping 18%. By 1960, it had dropped to 6%, and in 2012, it bottomed out at 3.7%. It’s since crept back up to 5.9%, for reasons that aren’t fully understood. These deaths usually result from painful conditions like disease, infection, or organ failure, so a lower mortality rate suggests that at least some aspects of welfare on the farm have improved.

To complicate matters even further, another welfare accounting method one might use is the amount of time chickens spend on farms. Thanks to faster growth, chickens now live much shorter lives before slaughter: 112 days in 1925, 63 in 1960, just 47 today. This means that we now only need half the chicken-days now to meet demand compared to how agriculture was in 1960. Whether you think shorter lives are a mercy or a loss depends on your ethical framework, but regardless, this is another kind of footprint that efficiency helps us shrink.

I’m not claiming right now that any one method of welfare accounting is the best one to use.3 But one of the lessons of We Are Eating the World, is that accounting methods really matter. And at the very least, we have to move past the overly simplistic heuristic that because industrial agriculture is both bad and efficient, efficiency itself must be harmful. Clearly, there’s more nuance here to understand.

Share

It’s helpful to consider some specific examples of how efficiency affects welfare. Take male chick culling: in the egg industry, male chicks are routinely killed because they don’t lay eggs, and are too scrawny for meat production. This isn’t because culling itself increases efficiency, but because the industry became more efficient by optimizing separately for egg-laying versus meat production. That split made male chicks in the egg sector effectively useless, turning culling into an unintended side effect. The distinction matters since it means the problem is solvable if the solution doesn’t sacrifice efficiency. Technologies like in-ovo sexing, which identify the chick’s sex before hatching, offer a humane fix that aligns with the economic incentives.

Another example is the leg and heart issues in fast-growing broilers mentioned earlier. These problems also weren’t deliberately engineered–they emerged as unintended consequences of selecting for rapid growth and breast yield. Again, if broilers can be bred to have stronger hearts, or if we can change husbandry practices to improve the leg issues (such as through better vaccination methods), we may be able to solve these issues without sacrificing productivity. Some might worry that solving these issues would enable producers to push for even more efficiency, but as we’ve seen from this analysis, that might not be such a bad thing.

Silver-Bullet Fantasies

The final chapter of Grunwald’s book is called “How to Save the World.” One might expect from the title a clear roadmap of how to reform the food system to meet all of our climate goals. But then one would likely be left unsatisfied, which Grunwald acknowledges. Essentially the advice boils down to “try harder, do better, and don’t make things worse.”

But the unsatisfying-ness is kind of the point–the temptation to think that there was such a clear and easy road map often leads us into what Grunwald calls “silver-bullet fantasies.” In reality, agriculture is messy and always will be. We will always need to eat, and eating will always use land that could have been used to store carbon. The first step is to make sure we use that land as efficiently as possible.

But efficiency can only take us so far. Industrial agriculture has real challenges, and to solve them we need cost-effective, abundance-preserving solutions. In our pursuit of these solutions we need to be smart about how we measure the impact, as sloppy accounting can point us in useless or actively counterproductive directions. This work is complicated, controversial, and error-prone, but there’s no replacement for rolling up your sleeves and actually doing it.

Reading the book, I kept thinking about our agricultural system in relation to our healthcare system. Modern healthcare is immeasurably better than it used to be, but there are still diseases that ruin lives and cause immense suffering. But each year, things get better, and once we’ve cured one horrible disease, we move on to the next one. Modern medicine will never be “solved,” because there will always be ways that life could be better. But at any given point, we do our best to make it as good as possible.

Grunwald’s vision strikes me as adopting a similar ethic towards the planet, and I think it applies to animal welfare as well. There will always be ways to improve the lives of animals, and ways to be more efficient with the animals we do raise. We should continually push on both welfare and efficiency until we’re raising only as many animals as necessary, and giving each of them the best life we can, better than what they’d have in nature. And once we’ve done that, we should push even further. Animal welfare may never be solved, but we can always try harder, do better, and work not to make things worse.

1. Who really captures the animal welfare premium?

I’ve previously claimed that animal welfare certifications don't really drive consumer behavior. Instead, consumers tend to rely on trusted brands like Vital Farms or Whole Foods for assurances around animal welfare. My primary evidence has been that certifiers typically charge producers very little for certification, while trusted brands can command substantial price premiums.

But this is only indirect evidence, and as far as I could find there is no rigorous, systematic study on consumer willingness-to-pay for animal welfare certifications. Existing research mostly relies on self-reported data, which can be unreliable. A better approach would involve controlled experiments comparing consumer behavior towards two similar products—one certified and one not—to measure actual willingness-to-pay. I'd love to see a thorough investigation into questions such as: How large is the welfare premium, and which parts of the supply chain currently capture it? What determines which part of the supply chain consumers look to for assurances around animal welfare?

The reason this is important is that certifiers and brands transmit different incentives up the supply chain. Brands often emphasize welfare aspects that are visually or narratively appealing, like outdoor access and naturalistic farming conditions. While animals on these farms might have better welfare than the current alternative, this branding promotes a pastoral view of humane farming that will struggle to meet global protein demand. Critical but less narratively appealing welfare factors such as humane slaughter practices, litter quality, ammonia concentration, or overall animal health often receive less attention. Certifiers can more effectively incentivize these less marketable welfare issues, and can be more objective as external auditors.

But if certifications don't meaningfully drive consumer behavior, these potential advantages become irrelevant. Strengthening third-party certifications could benefit everyone involved, enabling farmers to receive fair compensation for genuinely improving welfare, and allowing consumers to confidently purchase higher welfare products without needing to individually understand specific details about how farming works. Understanding the dynamics of certification is a critical first step toward this alignment.

Subscribe now

2. Should the government subsidize or regulate?

When it comes to government involvement in animal welfare, there are two possible approaches: governments can require farms to adopt better practices through regulation, or they can subsidize farms to adopt better practices. In the U.S., the government has taken a primarily regulatory approach, resulting in improved welfare but also generating strong industry opposition. This opposition threatens the long-term viability of these initiatives and makes animal welfare a much more contentious and acrimonious issue than it needs to be. I've previously argued that subsidization could be a more politically powerful and economically efficient way to improve welfare.

There are historical precedents for government subsidies supporting animal welfare outside the U.S. For instance, when France banned chick culling, it provided subsidies to the egg supply chain to ease the transition.

I’d love to see more comprehensive research into the historical examples of government animal welfare subsidies and their effectiveness. How have these subsidies performed compared to regulatory approaches, particularly regarding political acceptance and effectiveness in achieving welfare improvements?

It would also be valuable to explore how to structure subsidies to achieve the best outcomes. Should the focus be on subsidizing supply-side changes or stimulating demand? Is it more impactful to subsidize proven technologies ready to scale or to support new, yet unproven innovations? In what contexts are market-shaping initiatives like advanced market commitments most effective?

3. How does innovation happen within animal agriculture?

I've previously argued that animal agriculture is particularly slow to adopt new innovations compared to other similar industries. This applies across the board, not just around welfare technologies. I presented a few hypotheses for why this might be: high levels of industry consolidation, significant capital requirements, and insufficient institutional support. However, I'd love to see a more systematic analysis of this question. How much do producers invest in R&D, and what practical impacts does this spending have across the supply chain? What role, if any, has venture capital played in driving innovation within animal agriculture?

I’m particularly interested to look at the role of academia in animal agriculture innovation. Beginning in the nineteenth century, the federal government created the land‑grant university system specifically to boost agricultural productivity and innovation. Even today, much of the innovation in animal agriculture appears to originate in these institutions rather than within producer R&D departments or external startups. I’d be interested in a study of how the land-grant university system has impacted innovation in animal agriculture. Is academia indeed more central to innovation in animal agriculture compared to other industries? How can research coming out of land‑grant universities be translated more effectively into practical on‑farm technologies? Considering potential cuts in government agricultural research funding under the current administration, what impacts might we anticipate for innovation in the sector?

The more animals there are, the more the economics favor group-level technologies. Their fixed costs are shared across many animals, whereas individual systems scale more linearly with head count. For many use cases, group-level technologies work perfectly well, but ultimately the outcomes that farmers care about are traits of individuals, whether better welfare or higher productivity. Group-level technology necessarily targets proxies or aggregates of these traits, which often limits their effectiveness. In these cases, a frequent challenge is developing an individual-level technology that is more effective. I call this “individuation.”

Individual-level technologies are usually more complicated and expensive than group-level ones, but they’re often critical to fully solving challenges in animal agriculture. This is especially true for smaller animals, which are present in far higher numbers on farms.

In this post, we’ll discuss two examples of this phenomenon from aquaculture and poultry production.

Fish Stunning

One of the major animal welfare priorities in aquaculture right now is humane slaughter, the goal of which is to render animals unconscious and insensible to pain before they are killed. Humane slaughter for fish is particularly complicated, given how many different species we farm, each with a distinct size and biology. Today, many aquaculture operations still kill fish while they are conscious, either by suffocation or chilling in an ice bath. In the last few years, however, the aquaculture industry has begun adopting new humane slaughter technologies that are installed onboard harvesting vessels or directly in aquaculture facilities.

One of the most common humane slaughter methods for fish is electrical stunning. In this method, fish are pumped from their holding pens into stunning machines that deliver an electric shock designed to render them unconscious. In "in-water" stunning, fish move through a tube with submerged electrodes, completing an electric circuit that shocks them unconscious. In an alternative method, "dry" stunning, fish are deposited onto a waterless conveyor belt where electrodes positioned above and below the belt deliver the shock directly.

Electrical stunning is a group-level technology, since fish move through the tube or along the conveyor belt in a continuous mass. This means that salmon weighing up to 8 kg each can be stunned in a similar process to shrimps weighing 10 grams each.

However, recent scientific evidence suggests that electrical stunning may not reliably render certain species of fish unconscious. For shrimp and some smaller species, electrical stunning works effectively, often killing them instantly. But for larger fish, such as salmon and trout, there's a problematic tradeoff. If the electrical current is set too high, fillet quality suffers due to internal bruising, spinal fractures, or hemorrhages. Set too low, the shock may merely immobilize rather than truly stun the fish, leaving them potentially conscious before slaughter. It can be difficult to find an electrical intensity that reliably achieves humane stunning without compromising product quality.

A different technology, called percussive stunning, could work better. Instead of electrocution, fish are instead fed into individual channels where they are precisely struck on the head with a mechanical bolt, instantly rendering them unconscious. This is similar to captive-bolt stunning techniques commonly practiced for cattle, but adapted into a highly automated system capable of much higher throughput. Percussive stunning thus represents an individualized approach to humane fish slaughter.

Individually processing each fish also unlocks additional benefits. For instance, fish stunned percussively are often immediately bled in the same processing channel, greatly reducing residual blood in fillets. Immediate bleeding can improve freshness, shelf life, and overall fillet quality. Such product-quality benefits have already made percussive stunning common practice for high-value species like salmon and trout. But medium-sized fish like carp and sea bass, which historically have only been handled in bulk, have never before had access to these individualized processing advantages. Percussive stunning thus opens up entirely new quality opportunities for producers of these species.

However, it may be some time before we see percussive stunning used for these medium-sized fish. The cost of percussive stunning scales roughly in relationship to the number of fish processed, whereas electrical stunning scales in relation to the volume of fish, irrespective of the total number of individuals. Consequently, the smaller the species, the greater the cost advantages of electrical stunning over percussive stunning. Some technological innovation will therefore be necessary before percussive stunning can become economically viable for these smaller, lower-value species.

Precision Livestock Farming

This tension between welfare, economics, and individuation isn't limited to fish stunning; a similar dynamic occurs in Precision Livestock Farming (PLF), a set of technologies designed to closely track animal productivity, health, and welfare. The exact technologies employed for PLF vary substantially depending on the species involved.

For larger and more economically valuable animals like cows or pigs, PLF usually involves highly individualized methods. For example, many dairy farms now place RFID ear-tags or rumen boluses on cows. These individual-level devices track each animal’s body temperature, rumination patterns, and activity levels, quickly alerting the farmer if an animal is ill, stressed, or ready for breeding. The farmer can then intervene specifically with that animal, treating it individually based on its unique needs.

The economics of individuation are relatively straightforward in these cases: a single dairy cow might produce thousands of dollars of milk annually, making it easily worthwhile to invest $20–$40 in an ear-tag sensor to maintain optimal productivity and welfare. But the economic calculus changes for smaller animals, especially poultry. Chickens might be worth only a few dollars at harvest, meaning individualized technologies need to be incredibly cheap to be economically feasible. At such low per-animal margins, individual RFID tags or wearable sensors are financially out of reach.

Instead, PLF for poultry tends to be more group-oriented, relying on tools like overhead cameras, environmental sensors, and microphones installed throughout barns. These group-level technologies continuously monitor conditions such as temperature, humidity, air quality, bird distribution, and flock noise patterns. They then aggregate this information into barn-wide metrics, alerting farmers if conditions deviate significantly from expected patterns.

Such approaches have been moderately successful at flagging issues like disease outbreaks, temperature stress, or feed problems, but fundamentally have a lower value proposition than PLF technologies for larger animals. This is why PLF is significantly more common for cows than for chickens.

In order for PLF to become widespread for chickens, it must be individualized in an extremely cheap way, on the order of one cent or less per bird. At this price level, anything involving hardware for individual chickens likely becomes infeasible. Instead, the most likely way PLF establishes a significant value proposition for poultry producers is if individual tracking and monitoring can be done completely via computer vision and software. It’s clearly within current AI capabilities to track individuals within a large barn, and if a single system can monitor thousands or even millions of birds at once, the cost of PLF techniques could fall substantially.1

Technology Over Pastoralism

Individualized technologies usually rely on heavy automation, mechanization, and economies of scale—features that contrast sharply with the pastoral ideals traditionally associated with humane agriculture. Modern visions of humane farming often evoke a lone herdsman and a small flock on open pasture. These ideals emphasize the farmer’s direct involvement in husbandry, the animal’s role in regenerating the land, and the importance of a naturalistic environment.

This pastoral approach works up to a point because a single farmer can keep track of a few hundred or even thousands of animals. Scale it to fifty-thousand broilers or five-million tilapia, however, and the arithmetic collapses. The only way to guarantee positive welfare at these numbers is by using technology to scale the capabilities of individualized husbandry.

The emphasis on pastoral methods has emerged because society tends to focus on big, high-value mammals. Beef and dairy are much larger industries than poultry or seafood, and account for an outsized share of the climate impacts of animal protein production. Also, the cattle rancher holds a special place in the American psyche and self-conception.

But smaller low-margin animals make up an overwhelming majority of animals that we consume. There are fewer than 100 million cows in the US, but almost 10 billion chickens—two orders of magnitude more. On top of that, most of the growth in animal protein consumption is coming from chicken and fish.

Therefore, if we want to meet the growing demand for animal protein while building an agricultural system aligned with our values, we can’t simply apply the pastoral logic that works for grass-fed beef or smallholder dairies across all of animal agriculture. We need a new technocratic vision of humane farming that leans into automation, engineering, robotics, and economies of scale to guarantee positive welfare for billions of individual animals at once.

Progress Studies is something like a gratitude exercise. Most of us live lives of remarkable abundance, yet this simple fact can be surprisingly hard to recognize. Progress Studies seeks to restore an appreciation for the ways in which our lives and civilization have improved dramatically relative to the past. Then, it seeks to understand the mechanisms behind these advances to help ensure the future is even more abundant.

Because progress is multifaceted, Progress Studies has a remarkably diverse and broad ranging set of interests. Yet animal welfare has sometimes sat awkwardly within the Progress Studies portfolio. The average life of a farm animal has gotten worse in the last few decades, running counter to the central pro-progress narrative of Progress Studies, and some thinkers have struggled to integrate it into their broader worldview. For example, Tyler Cowen, one of the intellectual fathers of the Progress Studies movement, acknowledges in his broad philosophical treatise Stubborn Attachments that his theory “cannot resolve long-standing disputes over animal welfare and animal rights.” Noah Smith says that animal welfare is something that “tempers” his general progress-oriented mindset.

However, I would argue that the philosophy of techno-optimism, which Progress Studies already applies to areas like climate change, is already well suited to deal gracefully with the issue of animal welfare. In particular, agricultural technologies that directly improve welfare while keeping food prices affordable represent a promising yet neglected path to meet global protein demand while staying true to our morals.

Technology eliminates tradeoffs

The fundamental challenge of animal welfare is that human preferences are multifaceted: we want better animal welfare, but we also want cheap, abundant food. During the industrialization of the last decade when we’ve proven that life doesn’t have to be defined by poverty, hunger, and material scarcity, the latter preference has tended to win out.

However, now that there are animal products in every grocery store, and meat is a part of the daily diet of most Americans, we have the breathing room to look back and ask if we’re fully satisfied with what we’ve built. And many of the common agricultural practices that contribute to the affordability of food strike most people as clearly unacceptable.1 But the tradeoff remains: many believe that meeting our own standards for humane treatment means letting food prices rise unacceptably, or forgoing meat altogether.

In its pursuit to understand the mechanism of human progress, Progress Studies has a useful lesson: If a “solution” to a problem involves everyone suddenly changing their preferences, then it’s not a real solution. A real solution gives us everything that we want. In this case, the only way to meet our preferences for both food abundance and animal welfare is to use technology to dissolve the tradeoff entirely. This philosophy can be described as “techno-optimism.”

Noah Smith illustrates this philosophy through a thought experiment involving a group of people that are sustained by the seeds of a grove of fruit trees. First, they develop technology to cook these seeds, leading to a period of growth and abundance. However, this prevents them from re-using the seeds to grow more trees, and eventually their grove dies out. One approach, which Noah calls “degrowth,” is to simply not cook the seeds, and revert back to the way things were. This would be the equivalent of abstaining from meat consumption. The other approach, techno-optimism, is to develop technology to intensively cultivate the seeds so they can grow more seeds while still being able to cook them.

Progress Studies’ techno-optimist stance is most clearly seen in how it deals with climate change. Humans want a sustainable climate, but also the economic benefits gained from fossil fuels. Here, Progress Studies favors dissolving the tradeoff with technologies like cheap solar panels, shale gas, nuclear power, EVs, and lithium-ion batteries. The techno-optimist perspective on climate change is vindicated by the fact that per capita carbon emissions in the US now have dropped below WWI levels (yes, that’s true!).

The analogies between animal welfare and climate change run deep. Both deal with the unintended negative side effects of human economic progress. Both deal with commoditized, capital intensive parts of our economy. And both involve systems that every person interacts with multiple times a day—energy and food. Given the structural similarities, the solutions should be similar as well. Techno-optimism is suited to both precisely because it takes human preferences seriously and gives us everything that we want.

How to make progress on animal welfare

Applications of techno-optimism to animal welfare have historically been focused on alternative proteins like plant-based and cell-cultivated meat.2 The promise of alternative proteins has been to shift the method of producing meat to one where no animals are involved at all, while continuing to meet human preferences around meat consumption.

But more recently, lagging sales numbers for plant-based meat and concerns over the scalability of cultivated meat have cast into doubt the long term potential of the sector. While it’s natural for all new technologies to go through cycles of hype and disillusionment, it’s a mistake for alternative proteins to be the only bet that Progress Studies makes in the realm of animal welfare.

Fortunately, there’s another techno-optimist approach: advancing agricultural technologies that make it cheap for farmers and agribusiness to solve the biggest welfare challenges on farms.

In-ovo sexing is a recent example of the power of such technologies. One of the reasons chicken meat and eggs are so cheap today is because we’ve developed specialized breeds for each purpose. “Broiler” chickens are optimized to quickly and efficiently convert feed to meat, and “layer” chickens are optimized to lay eggs as efficiently as possible. However, one unintended consequence is that male chicks of the layer breed serve no economic purpose and are therefore killed immediately after hatching. In a practice that’s extremely unpopular among consumers who know about it, six billion day-old male chicks are killed each year in the global egg industry.

In-ovo sexing allows egg producers to use advanced biotechnology to identify which eggs will hatch male and which will hatch female. Male eggs can be removed and destroyed before they can feel pain, leaving only females to hatch. This technology is now widely available in Europe, and more recently launched in the US, with the first eggs expected to hit shelves in the summer.

Similar opportunities exist throughout animal agriculture. Precision livestock farming, for example, can give us the ability to monitor the health and welfare of millions of individual animals at once. Fish stunning equipment can allow aquaculture farmers to ensure each fish is unconscious before slaughter. And new vaccines can allow us to treat health issues that currently kill hundreds of millions of animals per year.

The proliferation of such technologies into animal agriculture won’t happen automatically, but will require continual focus and effort. We need more scientists and engineers to conduct R&D around cost-effective ways to solve big welfare challenges, and we need more funding for their research. We need to create stronger market incentives for the commercialization and scale up these technologies through government subsidies, advanced market commitments, and innovation prizes. We need consumers to support innovative companies with their wallets, and we need a transparent supply chain that better transmits willingness-to-pay signals so that producers can be properly compensated for doing things better.

The details of this perspective still need fleshing out, and one goal of this newsletter is to develop its intellectual foundations more fully. But in the long term, I believe that agricultural technologies have the potential to transform the way we do animal husbandry and ideally give animals a quality of life that’s impossible even in nature, just as progress has made our quality of life unimaginable to our forebears. Our task now is to thoughtfully apply the intellectual tools that created progress for humans to create steady, meaningful progress for animals as well.

• Due to Highly Pathogenic Avian Flu (HPAI), 168 million chickens have been culled and the price of eggs reached historical highs. This crisis can be stopped if the USDA permits the use of an HPAI vaccine to protect chickens, American consumers, and farmers. I was on a panel hosted by the conservative think tank American Enterprise Institute for a post-mortem on the bird flu, the resulting price of eggs, and how to prevent this from happening in the future. The panel was streamed on C-SPAN, and covered on the trade websites Agri-Pulse, WATT, and The Fence Post.
  

• I also recently published an op-ed in Agri-Pulse on the importance of vaccination. Worryingly, I wrote that this might just be the beginning of sky-high egg prices: "Many experts now believe the virus has become endemic, with year-round detections in wild birds suggesting this pattern of price spikes will continue indefinitely without vaccination. Additionally, the longer bird flu is allowed to circulate in livestock populations, the greater the risk of a human pandemic."

• In January, we published an annual report reviewing the development of in-ovo sexing, followed by an update in April. There has been a lot of big news:

  • The first in-ovo sexed chicks were hatched on U.S. soil.

  • Following Germany and France’s bans, the Netherlands plans to end male chick culling by 2026.

  • Walmart included in-ovo sexing as an important focus area for its egg suppliers.

  • There was a 40% increase in the number of in-ovo sexing machines worldwide in 2024.

  • The market continued to expand worldwide!

• Perdue Farms became the first major US poultry producer to adopt on-farm hatching which could be one small step towards the chicken farm of the future.

• In January, I gave a talk to 200 poultry executives at the poultry industry’s largest annual conference (IPPE) on "3 Underdiscussed Tends in the Poultry Industry". The talk focused on:

  • 1) The mysterious decline in productivity across the US broiler sector.

  • 2) The broader supply chain implications of technologies like in-ovo sexing, on-farm hatching, and in-ovo vaccination for the transition to the hatchery of the future.

  • 3) The importance of animal welfare for American consumers and the role that technology plays in meeting these demands.

• We launched a report on Brazilian consumer attitudes towards in-ovo sexing. In a nutshell, Brazilian's are willing to pay substantially more for eggs that come from companies using in-ovo sexing. For Portuguese speakers, here’s some of the coverage that it’s been getting.

• We recently put together a report comparing the economics of fish harvesting with electric versus percussive stunning technology. Overall, we found that percussive stunning has significant economic benefits for salmonid farmers, and we identified a number of technological improvements needed to adapt the technology for smaller fish like carp. Please reach out if you’re interested in viewing the report.

• In January, Niko McCarty wrote an essay on in-ovo sexing for The New Biology.

• A New York Times op-ed on in-ovo sexing was released. Sy Montgomery writes that eggs from companies utilizing in-ovo sexing (NestFresh and Kipster) will become available in the early summer. Expect a lot more coverage among major American news services as summer approaches. It’s important to note that the additional cost of per egg should not exceed 1 cent, less than Montgomery estimates.

  

• A new vaccination method has been shown to decrease lameness among boiler chickens by 50%! As I wrote in a note, "This study is super exciting because: 1) The effect size is very large. Other cheap interventions for lameness might only reduce it by a few percentage points 2) Vaccines are extremely cheap, and broilers are already vaccinated for other diseases like Marek's." Read the full study here.

Innovate Animal Ag, the think tank behind The Optimist’s Barn, is hiring! And if you read this newsletter, that’s a pretty good indicator of potentially being a fit. To learn more, visit our Careers page.

This post is the second in a series where we set aside the economic and engineering constraints that drive husbandry today, and instead imagine what the most humane form of husbandry might look like in a future of immense abundance. If you haven’t read Part 1, I’d recommend starting there. Though this vision may still be far off, it's important to define what it could look like so that we can help steer things in the right direction.

Automated Husbandry

In Part 1, I described an integrated supply chain without live transportation, where chickens hatch, grow, and are slaughtered within a single high-tech facility. Flocks of chickens live in small, carefully constructed rooms, each doubling as a controlled-atmosphere stunning chamber. But the critical question remains: how can we systematically ensure humane treatment for each individual chicken throughout its life?

At the heart of humane poultry farming in the future will be comprehensive automation driven by artificial intelligence. Advanced AI technologies could automate nearly every aspect of husbandry, providing each chicken a level of individualized care currently impossible at scale.

Cameras could continually track every single individual in the room and assess their health in real time, immediately flagging any welfare concerns. Automated food and water dispensers could also deliver precisely optimized diets or administer targeted medications based on real-time biometrics.

AI-powered robotic caregivers would perform veterinary-level interventions, replicating, and eventually surpassing, the abilities of human farmers. Today, human involvement remains important because effective husbandry relies on craft and intuition developed through years of experience. Farmers who spend years caring for chickens become finely attuned to the subtle cues of their flock, recognizing specific movements or vocalizations that indicate a flock's needs.

However, recent advancements in AI have made it clear that husbandry could in theory be fully automated. Current LLMs can already outperform human doctors for human medical diagnostics, and frontier labs are rapidly advancing AI agents capable of the reasoning over time required to manage a flock from hatch to slaughter. And unlike human caregivers, AI systems won't be limited by the number of animals they can effectively monitor. Additionally, such monitoring could also flow into an independent welfare verification scheme.

Further, having humans physically onsite itself has a number of downsides. Even with stringent biosecurity protocols, human workers remain a primary vector for pathogens entering poultry barns. Human caregivers, despite their best intentions, can be susceptible to fatigue, inconsistency, and error—risks that are minimized with the consistent, precise performance of AI-driven robotic caregivers. Chickens are naturally stressed when they face unfamiliar stimuli like human handlers, but robotic caregivers would represent a constant, predictable presence to which chickens could readily adapt.

Ultimately, when husbandry can be completely automated, removing humans completely will have benefits to both animal welfare and economics.

However, some health issues may inevitably arise. When they do, having robots capable of humane euthanasia will be essential, especially if the health issue risks spreading to other birds. Advanced bioanalytics could proactively detect subtle indicators suggesting a bird is on a trajectory toward irreversible health decline, enabling euthanasia before the animal experiences significant discomfort or negative welfare. In these situations, an automated euthanasia robot could gently approach the affected chicken, calmly transfer it into a container with a nitrogen atmosphere for painless euthanasia. This capability is already within reach of current in-barn robotics companies.

While AI will drive most capabilities, increased societal wealth and technological progress will also facilitate crucial facility improvements. For example, powerful ventilation systems could maintain ammonia and particulate levels at comfortable, healthy thresholds. Far-UVC lighting would ensure each room remains free of pathogens, drastically reducing disease risks. Low stocking densities would give chickens ample space to move freely, eliminating stress from overcrowding. And thoughtfully designed environments—including regularly cleaned dirt floors for natural dust bathing, comfortable perches for resting, pecking substrates for satisfying foraging instincts, and sheltered areas to seek privacy and perform nesting behaviors—would allow chickens to express their full range of natural behaviors.

Subscribe now

To evaluate whether this system delivers the kind of husbandry we’d feel proud to support, we can turn to the Five Freedoms framework—a widely respected model of animal welfare that captures the core conditions necessary for animals to live well.

1. Freedom from hunger and thirst: Food and water would always be plentiful, with chickens consistently healthy enough to access them.

2. Freedom from discomfort: The environment would be spacious, stable, and protected from adverse weather, predators, and pathogens.

3. Freedom from pain, injury, and disease: Health issues would be treated on an individual basis and any bird with irreversible health issues would be humanely euthanized before experiencing negative welfare.

4. Freedom to express normal behavior: Enriched, stimulating environments would enable natural social interactions and behaviors like dust bathing, foraging, and nesting.

5. Freedom from fear and distress: Without the unpredictable presence of human handlers or sudden environmental disruptions, chickens would rarely encounter frightening stimuli, significantly reducing stress and anxiety.

When every aspect of a chicken’s life is fully automated, each individual could experience a calm, stable, and responsive environment from the day it hatches. Over the course of its life, every need would be anticipated and met. Then, one day, it would fall asleep and simply not wake up.

Technology over pastoralism

This vision of humane husbandry runs counter to many contemporary trends in animal agriculture. Current welfare initiatives often emphasize pastoral ideals such as outdoor access, smaller flock sizes, and more direct human involvement in care decisions. But these are only proxies for the actual thing we care about—ensuring the best possible welfare for each individual animal. In the future, more technological approaches will allow us to directly provide better welfare more effectively and efficiently.

For example, outdoor access is valued because it gives chickens access to sunlight, fresh air, vegetation, and dirt for dust bathing. But thoughtfully engineered indoor environments with strong ventilation, windows, and enrichment can supply all these benefits without exposing birds to the risks of predation, harsh weather, or diseases carried by wild animals.

Innovative modern-day companies like Kipster already demonstrate how outdoor access is unnecessary for good welfare. Kipster's indoor egg farms provide some of the highest-quality living conditions in the industry, which they show off via livestreams viewable at any time. On their farms, sunlight streams in through ceiling windows, and chickens have access to enrichments through “indoor gardens.”1

Even though “pasture-raised” eggs are commonly considered to be the highest welfare standard in the egg industry, Kipster doesn’t use this label because technically their chickens don’t have access to pasture. In fact, Kipster argues that traditional pasture systems can introduce stress, as chickens often fear exposure to predators outdoors.

The pastoral ideal for animal agriculture assumes that nature and natural environments inherently provide the best welfare. But for humans, it’s been our use of technology to move away from the state of nature that has enabled the greatest flourishing. Kipster has begun to demonstrate how this may be true for animals as well. As technology unlocks powerful new capabilities, and growing societal wealth makes them broadly affordable, we should actively embrace these advancements. Rather than clinging to outdated pastoral ideals, we should leverage human ingenuity to envision how we can offer animals better lives than what’s possible in nature.

In the next post in this series, we’ll wrap up the poultry industry by going deeper into how we can raise chickens that aren’t used for meat, but eggs. This includes both layers and breeders (the parents of broilers or layers). Stay tuned!

Innovate Animal Ag, the think tank behind The Optimist’s Barn, is hiring! And if you read this newsletter, that’s a pretty good indicator of potentially being a fit. To learn more, visit our Careers page.

These metrics span every stage of the broiler supply chain:

1. Breeder farms, where “broiler breeders” lay fertilized eggs which eventually become birds used for meat. Here, the relevant metric is the number of eggs laid per year by each breeder.

2. Hatcheries, where these fertilized “hatching eggs” are incubated and hatched. Here, the metric is the percentage of hatching eggs that yield viable chicks, called “Hatchability.”

3. Grow-out farms, where “broilers” are raised to their slaughter weight. Here, the metric is the percentage of those chicks that die before reaching market, called “Mortality.”

All three metrics are reported or derived from USDA data. They generally improved until about 2012, then reversed course and have been sliding ever since.

Mortality, for instance, started at 18% in 1925 and then fell for almost 100 years to a low of 3.7% in 2012. Then, it started steadily increasing to 5.93% where it now sits today. Hatchability followed a similar trend, roughly improving through the late 1990s and 2000s. But starting around 2012, after peaking around 85%, the trailing twelve-month average started steadily declining, reaching 79% in 2024. The number of eggs per broiler breeder has been a little more chaotic, and has been declining since around 2002. But 2012 saw a sharp inflection where it fell to 90% of 2002 levels in just 4 years, then further declined from there.

Taken together, these trends are causing the breeder flock to balloon relative to the number of broilers. Since 2012, broiler production has increased by 8%, while the breeder flock expanded by 20%. At the same time, the increase in mortality mans that more broilers are lost before reaching slaughter. In other words, significantly more chickens are now needed to produce the same amount of meat.

This decrease in efficiency hasn’t had a major effect on chicken meat prices, potentially because the more economically important metric is feed efficiency–the number of calories of feed it takes to generate a calorie of meat. This metric hasn’t followed the same pattern, since it has declined every year since 1925. However, needing more birds to create the same amount of meat would clearly have negative economic effects. It is also bad for sustainability and animal welfare. So why is it happening?

It’s not just antibiotics

Each of these metrics are multifaceted. It's easy to come up with 20 hypotheses for things that might cause each individual trend. But, all these trends started to reverse around 2012, strongly suggesting a single underlying cause. Otherwise, it seems like too big a coincidence that every link in the broiler chain would start trending downward at once.

Something else that happened in 2012 was that the FDA heavily restricted the use of antibiotics in animal agriculture. At first glance, this seemed like the obvious explanation.

Historically, antibiotics were used on farms for three reasons: treating active disease, preventing future disease, and, most controversially, boosting growth and feed efficiency. Because farming environments tend to be fairly dirty, farmers discovered that low, persistent doses of antibiotics helped chickens grow faster and more efficiently, which had a direct positive impact on the farmer’s bottom line.

As a result, using antibiotics for growth promotion became commonplace on broiler farms. Naturally, this sparked concern about antibiotic resistance, especially as there started to be specific instances of antibiotic resistant bacteria traced back to animal agriculture. The FDA’s response in 2012 was to effectively disallow the use of medically-important antibiotics for growth promotion, and require that antibiotics be administered under veterinary supervision only to treat or prevent disease. This led to a significant decrease in the amount of antibiotics used on US poultry farms, as well an increase in the popularity of marketing terms like “antibiotic-free.”

Share

Initially, it seemed like this could be an open and shut case. Decreasing antibiotics can easily lead to an increase in broiler mortality - if companies focusing on antibiotic-free production no longer administer antibiotics when birds get sick, that would lead to more birds dying on average. It can also easily explain a decrease in eggs laid per breeder, if you assume that general bird health is a major input to the number of eggs laid. Explaining the decrease in hatchability through antibiotics is a little trickier, but it can work - fertilized eggs are sometimes contaminated which causes the embryo to die, or sometimes an infection can cause a hatching egg to explode in an incubator (called a “banger”), contaminating the entire batch. If reduced antibiotic usage was somehow increasing bacterial load on hatching eggs, that could lead to the decrease in hatchability we’ve been seeing.

But upon further research, this hypothesis started to look less good. Breeder farms had little incentive to give sub-therapeutic doses of antibiotics in the first place. Breeders are raised for their eggs, not meat, so optimizing their growth rate doesn’t yield the same economic benefit. That leaves fewer opportunities for antibiotic reduction to affect breeder-level metrics like egg production. Additionally, antibiotic use at the breeder level doesn’t impact “antibiotic-free” labeling, so there’s no marketing-driven reason to cut back on antibiotics on breeder farms.

There’s relatively little data on what happens on breeder farms, which makes it tough to say for sure what did or didn’t change. However, egg-laying hens could provide a useful comparison. Like broiler breeders, layers are used for egg production, so their incentives more closely resemble breeders than broilers raised for meat. The data for layers suggests that eggs per layer per year have been continually increasing over time. Additionally, layers receive significantly less antibiotics than broilers. As of 2020, broilers receive on average 0.1 grams of antibiotics over the course of their 4-6 week lifespan, whereas layers receive on average 0.03 grams over the course of their 18-month lifespan.

The FDA’s 2012 antibiotic ruling applies to the egg industry as well, yet there’s been no notable dip in layer or layer breeder performance that matches the downturn seen in broiler breeders. If the regulatory change were the key driver of breeder inefficiency, we would expect a parallel effect in layers—but we don’t see it.

The antibiotic hypothesis was difficult to give up, given how beautifully it seems to line up with the data. Especially given that further statistical analysis suggested that 77% of the mortality change in broilers could be explained by reductions in antibiotic usage that we’ve in fact seen since 2012. Additionally, the prevalence of Salmonella in chicken meat seemed to be increasing in strong correlation with both mortality and the percentage of chickens raised without antibiotics.

But mechanistically, a reduction in antibiotics can’t as neatly explain the changes on the breeder farm and hatchery levels. For example, hatching eggs are generally heavily disinfected as they enter the hatchery which could serve as an equalizing factor for bacterial load. There are other ways in which pathogens can persist inside a hatching egg, but they’re less common and tend to have a smaller effect.

An unsolved mystery

So it’s clear that antibiotics aren’t the full story. However, I’m fairly confident that they did play some role in the increase in broiler mortality since 2012. First, the statistical correlation between reduced antibiotic use and higher mortality is strong. Second, the causal link is straightforward: fewer antibiotics mean more birds succumb to illness, which inevitably pushes mortality up.

That leaves the declines in hatchability and eggs laid per breeder still unexplained. These trends appear to stem from at least one other factor—one that could also in theory amplify broiler mortality. Both forces would have begun taking effect around the same time, perhaps coincidentally. This picture isn’t as simple or beautiful as a single underlying factor, but sometimes the world doesn’t want to be explained by simple hypotheses.

Once we’re no longer looking for a factor that can explain all three trends, there are suddenly many other possible causes. To list a few:

• Over time, broiler genetics have been increasingly optimized for faster growth and larger size, traits that directly reduce the cost of producing chicken. Since broiler breeders necessarily share genetics with broilers, they are also affected by these genetic changes. But unlike broilers, breeders live much longer, which can magnify health problems tied to growth-focused traits and reduce their ability to lay eggs. Another side effect could be that larger breeders may no longer be able to comfortably lay eggs in the nest boxes that they’ve historically used. This could cause more eggs to be laid on the floor of the barn, where bacterial loads are significantly higher, causing reduced hatchability.

• Macroeconomic factors may have also had subtle effects. For example, if demand for chicken meat increased faster than anticipated, breeder farmers might have kept flocks for longer in order to meet demand even though older broiler breeders are less efficient and have generally lower hatchability. Or, if other factors discouraged reinvestment in new facilities or equipment, aging infrastructure could have driven further efficiency problems throughout the supply chain.

• There could be specific pathogens that have been negatively affecting breeder health. One theory blames Enterococcus Faecalis, a microbe that is commonly found in poultry environments but is often dismissed as harmless However, more recent research has found the bacteria in dead embryos and infertile eggs, and has suggested that it can cause significant embryo loss in infected hatching eggs.

I’m not sure which of these hypotheses is correct, or whether the underlying cause is a mix of different factors. There’s a lot of complexity and uncertainty here, and I’d be very interested in seeing more research devoted to this topic.

What did we miss? Do you have your own hypotheses? Let me know in the comments!

Innovate Animal Ag, the think tank behind The Optimist’s Barn, is hiring! And if you read this newsletter, that’s a pretty good indicator of potentially being a fit. We’re not looking for folks with specific types of experience, we want exceptional people that are deeply passionate about our mission and theory of change.

To learn more, visit our Careers page.

No matter how much this growth changes society, humans will (probably?) stay the same. We’ll still want to raise animals for food. But we’ll also be able to afford vastly more accommodations for things like animal welfare.

One of the challenges of this newsletters’ project of applying a techno-optimist lens to animal agriculture is that it can be hard to envision the type of future we’re building towards. For sustainability, we can imagine a world where solar panels, wind turbines, and nuclear power plants provide limitless clean energy, and the electrification of every sector allows us to turn this energy into limitless productivity. That world isn’t unimaginably different from our own. What is the equivalent vision for animal farming?

The short answer is that I don’t exactly know. But I think we can and should take a stab at starting to envision it.

This is the first part of a multi-part series that tries to imagine how animal husbandry might work in a world where there are vastly more resources to throw at every problem. The ideas I discuss in this series are not meant to be prescriptive, just a first stab. Ideally, volumes of research would be devoted to this question, as has been done to map out the clean energy transition.

The practical implications of this exercise are necessarily limited. In a world of truly transformational AI, there’s no way to know what kind of scientific advancements will be made, meaning that whatever I say here will necessarily be too conservative and unambitious. And in the short term economics will continue to be a limiting factor on achieving this strange and exciting future.

However, I hope this exercise can help justify that it is, in theory, possible to build an agricultural system that can meet global protein demand while also being in line with our values. It can also serve as a roadmap to help us understand what kind of capabilities we need to start building.

We'll first turn attention towards chickens that we use for meat (called broilers), and in particular what a supply chain might look like that's optimized for welfare. Today's chicken supply chain is fragmented into multiple specialized facilities, a structure shaped largely by historical economic incentives. In a world of greater wealth, these incentives will be less important, allowing us to imagine a system where chickens live their entire lives in one location, to the benefit of their welfare.

Why chickens move around

A broiler chicken will generally visit three facilities over the course of its life, each with distinct requirements around equipment, facility infrastructure, and staff expertise.

First, it begins as an embryo in an egg that's incubated and hatched at a hatchery. These facilities run large incubators which maintain precisely controlled conditions for egg development. They also use specialized equipment for vaccination, sorting, and sexing of newly hatched chicks.

Then, the chick is transported to a farm where it will spend 4-6 weeks rapidly gaining weight. Farms need ample floor space since broilers are generally raised on the ground. They also need feeding, watering, and climate control systems. Farmers themselves must be skilled in husbandry and supported by veterinary expertise.

Finally, the mature bird is transported to a slaughter facility where it's killed and processed into marketable meat. This specialized facility contains industrial-scale machinery for stunning and killing birds, along with production lines where carcasses are hung on shackles to be bled, cleaned, cut, and packaged.

Historically, these three steps fragmented as the poultry industry industrialized. Early on, smaller independent companies specialized in a particular stage because each required unique equipment, facility infrastructure, and expertise. Over time, the industry vertically integrated, with large companies now controlling multiple steps in the supply chain, but the specialized facility model remained.

A key reason why these steps are separated is their different economies of scale. Hatcheries and slaughter plants use heavy machinery to process huge volumes, which they can do because eggs and carcasses are relatively easy to work with—they’re inanimate and uniformly shaped, making them amenable to automation. Farms, on the other hand, are limited in their throughput because they need substantial floor space to rear live birds, and there’s a limit to the extent they can speed up the chicken’s biological process of gaining weight. As a result, relatively few hatcheries supply many farms, which then supply relatively few slaughter plants. This fragmentation allows hatcheries and slaughter plants to maximize their efficiency through scale without being constrained by farming's inherent scaling limitations

While the structure of the current supply chain therefore makes economic sense, every move between facilities involves inherent welfare challenges. Each transition involves some degree of handling, which is always stressful for birds. Additionally, transport trucks generally lack infrastructure like food, water, and climate control, so transportation increases the risk of mortality and disease. In a supply chain optimized for welfare rather than economics, transportation would ideally be completely eliminated. Fortunately, with some existing technologies and a few new capabilities, this could be possible in the future.

On-farm hatching and slaughter

To eliminate transportation, a single facility must be able to accommodate hatching, rearing, and slaughter. In parts of Europe, and now in some US operations like Perdue, part of this is already beginning to happen with on-farm hatching. Instead of hatching chicks in a central hatchery, eggs are delivered to the farm shortly before they hatch, then laid out on or near the ground. The chicks then hatch in the same environment where they’ll live, avoiding the noisy, dusty, and stressful conditions associated with the hatchery and transporting vehicle. Additionally, chicks have immediate access to food and water, which is better for gut health and welfare. Improved gut health and lower stress levels, combined with a reduced exposure to pathogens, means that significantly fewer antibiotics are needed for on-farm hatched chicks. This is a proven practice that is already becoming more common around the globe.

But what about on-farm slaughter? If it were possible, chickens could remain in one place for their whole life. The chicken meat industry has already started to explore mobile slaughter units, which show promise for both economic and welfare benefits. But in a future with far greater resources, we could take a more ambitious step: merging farm and slaughter operations into a single, fully integrated facility. This would be more complex and require some technological advancement, but the purpose of this thought exercise is to imagine what's possible if we aren’t bound by such practicalities.

When it comes to animal welfare during slaughter, the gold standard is using inert gasses like nitrogen or argon. If done right, these methods can be as close to undetectable by the chickens as possible, especially if no handling is involved. Currently, controlled atmosphere stunning (CAS) is commonly considered to be the most humane way to slaughter chickens at scale. With this method, chickens are put in an airtight container where inert gas is gradually added until the animals are unconscious. Then, carcasses are manually moved from the CAS container to the shackle line.

While CAS is a big welfare improvement over other methods, birds still need to be caught, put into crates, transported to the slaughter plant, and then loaded into the CAS machine. This process inherently causes stress, welfare issues, and mortality.

But what if the room where the chickens lived was itself a controlled atmospheric stunning chamber, and the slaughter plant was on the same site as the farm? One could imagine a setup where chicks lived their entire lives in one room gaining weight after being on-farm hatched. Then, when it was time for slaughter, the nitrogen valves turn on, and within a few minutes the chickens are unconscious, and their carcasses are shackled and moved to the slaughter facility for further processing.

One of the engineering challenges of this setup would be minimizing the time in between gassing and processing. Generally, birds need to be bled out as quickly as possible after gassing renders them unconscious. If they aren’t bled out, one of two things might happen: birds that merely passed out and were not killed from the gas might wake up, or for the birds that were killed, blood might coagulate in their vessels, decreasing meat quality.

The engineering challenge is therefore getting all of the birds that are unconscious scattered throughout a barn room onto a shackle line as quickly as possible. How exactly this could work I’m not sure, but hopefully the AIs of the future will have some ideas. Possibly, shackle lines would be directly accessible from the barn room and lead directly to the slaughter area, so that chickens could be quickly moved (probably with robots rather than humans) from the ground onto onto the shackles. One could imagine a facility design where rooms are laid out in a cube or grid, with a shackle line snaking through the design touching each room and eventually terminating at the slaughter plant.

Another engineering challenge would be sizing the farm and slaughter parts of the facility, and managing the flow between the two parts. The farming part would need to be significantly scaled up relative to the slaughter plant in order to produce enough birds to fill its capacity. However, if too many birds need to be slaughtered all at once, then there is a risk of overloading the slaughter throughput.

One way to solve this is to segregate chickens into flock sizes based on the processing capacity of the slaughtering operations. Each flock would then live separately in an airtight room within a larger facility. As a toy example, imagine a slaughter facility with a daily capacity of 100,000 head and a production cycle of 45 days. Birds could be divided into groups of 100,000, where each group is one day older than the last. This means that 4,500,000 birds would need to live on the farm at any given time (on the larger size of industrial farms today, but not unheard of). Each of these groups could then be further subdivided into multiple airtight rooms in order to time gassing with multiple slaughter shifts over the course of a single day.

With onsite slaughter using gas, birds could even be slaughtered while asleep, which is a level of humane slaughter that’s not currently possible given that handling will always wake birds up. Each room could be run on an artificial lighting schedule to allow for precise control over the birds’ sleep schedule.

The life of a chicken might then look like this: the chick hatches directly in a clean, calm environment; spends a few weeks roaming, eating, and socializing; then one day it goes to sleep and doesn’t wake up.

There’s obviously a lot more details to work out here, particularly regarding what the chickens’ lives are like inside these rooms. If each chicken lives a healthy life, free of significant pain, discomfort, or stress, and where they have the freedom to express natural behaviors, then I think we’ll have made a lot of progress towards the promise of humane husbandry at scale.

I think technology can help do this, which will be the topic of the next post of the series. After that, there are countless other contexts and animals to consider. For example, what happens to the parents of these chickens, called broiler breeders, which are raised in different types of environments with their own unique set of challenges? How about egg laying hens, fish (of which there are many types), pigs, cows, ducks, and turkeys?

There’s a lot left to go, so stay tuned.

Innovate Animal Ag, the think tank behind The Optimist’s Barn, is hiring! And if you read this newsletter, that’s a pretty good indicator of potentially being a fit. We’re not looking for folks with specific types of experience, we want exceptional people that are deeply passionate about our mission and theory of change.

To learn more, visit our Careers page.

The current state of play

Given how much egg prices are in the news, addressing bird flu was on top of the priority list of new Secretary of Agriculture Brooke Rollins. At the end of February, she unveiled her five point plan for fighting bird flu and lowering the price of eggs. One of the five points was “Explore Pathways toward Vaccines, Therapeutics, and Other Strategies for Protecting Egg Laying Chickens to Reduce Instances of Depopulation.” While her plan stopped short of approving vaccines for the egg, turkey, and dairy industries, it did allocate up to $100 million for vaccine research and included a commitment to addressing the trade-related challenges of a broader vaccination campaign.

More recently, Secretary Rollins retreated somewhat from vaccination, citing concerns over efficacy, and Robert F. Kennedy Jr, the new Secretary of Health and Human Services came out against vaccinating poultry against H5N1 on public health grounds.

USDA’s initial five point plan was applauded as a step in the right direction by industry groups, although they continue to advocate for full vaccine approval.

Trade wars

As I previously discussed, the biggest barrier to bird flu vaccination is international trade, specifically our ability to export chicken meat. The country that we export the most chicken meat to is Mexico (25%), followed by China (17%) and Canada (7%). These happen to be three countries that President Trump has recently announced tariffs on.

Indeed, chicken trade may be caught in the crosshairs as tensions between these countries rise. Canada’s retaliatory tariffs already include poultry meat (at the literal top of the list), as do China’s. Mexico hasn’t provided details on the products that will be subject to their retaliatory tariffs, but it might well include the $1B+ of poultry meat products that the US exports to Mexico each year.

It’s possible that if the trade wars escalate further and retaliatory tariffs reduce the flow of chicken meat from the US into Mexico, Canada, and China, then one barrier to bird flu vaccination could be removed. However, these escalations might also make it more difficult to undertake the negotiations necessary to protect US chicken meat exports if US egg producers are allowed to start vaccinating.

Given the volume of poultry products that flow from the US to Mexico, restricting US supply might be overly disruptive to the chicken meat market in Mexico. Imports from the US represent ~17% of the total chicken consumption in Mexico. If that supply were suddenly shut off, either because of a trade war or because of a Mexican response to a US H5N1 vaccination campaign, the price of chicken meat in Mexico could spike. Consider this in relation to the supply shock of eggs that’s occurred in the US, where we’ve seen a 9% decrease in egg supply lead to a 100% increase in egg prices. Chicken meat may have higher price elasticity than eggs because it can be substituted with other types of meat, but it’s reasonable to think that there would still be a significant increase in price for Mexican consumers.

In fact, US imports represent a significant fraction of the chicken meat market in many of the countries we export to. According to FAO data, 54% of US chicken exports go to countries where US chicken makes up more than 15% of domestic consumption. The possibility of price shocks in these countries makes it somewhat less likely that they would cut off American meat imports, whether because of bird flu vaccination or trade wars. However, in the quickly changing new international order, anything is possible, especially if countries are actively looking for opportunities to take more protectionist trade positions to bolster their domestic industries.

Subscribe now

A new Dutch vaccine trial

In more positive news, the Netherlands recently announced that it would be trialing vaccination on a single commercial egg farm. This trial is notable because the Netherlands is the world’s largest exporter of eggs, and this is the first vaccine trial in a major exporting country. Other countries like Mexico that currently vaccinate for bird flu generally aren’t major exporters, reducing the importance of trade implications. The eggs produced on the farm in question will only supply eggs domestically, but this trial could serve as an important indication of how the international community will react to vaccination campaigns in major exporting countries.

Notably, the “trial” is not to test efficacy - that was already completed to satisfaction in field tests last year. Rather the trial is to gauge the reaction of trade partners, as well as to gain experience with administration and surveillance of the vaccination program.

The vaccine in question, Innovax-ND-H5, works differently than the Zoetis vaccine that was given conditional approval in the US last month. It is a live recombinant vector vaccine that protects against H5 avian influenza viruses, and also Newcastle Disease and Marek’s, which egg producers already commonly vaccinate for. Unlike Zoetis’ killed-virus vaccine, the Innovax vaccine only has to be administered once, either in-ovo or on the first day of hatch.1

The ease of administration and the protection against other common diseases means that the Innovax vaccine has some clear practical advantages over the Zoetis vaccine. In recent congressional testimony, a representative from the egg industry emphasized the importance of these logistical considerations, since injecting each bird multiple times could be challenging across the entire US egg sector.

Such a trial could also be a sensible first step in a potential US vaccination program. In the same congressional testimony, the egg industry representative pointed out that there is a single egg farm in Hawai’i that could well suited for a trial given its geographic isolation from the rest of the US. A field trial on this farm could help test the efficacy of a vaccine in commercial farms without risking US trade relationships if eggs are only sold within the state.

All in all, a mass vaccination campaign in the US still may be a ways out, but this Dutch trial, as well as shifting trade realities could lower the barriers. I’ll continue to provide updates here as things develop.

The USDA's proactive approach comes at a critical time, as egg prices make headlines and consumers frequently encounter empty shelves at grocery stores. Since voters often view egg prices as a barometer for the health of the broader economy, addressing this issue has both practical and political importance. Politicians on both sides of the aisle have started talking about the issue, and given the role of inflation in the 2024 election, egg prices have the potential to be a major political win, or risk.

Undertaking a vaccination campaign of the US layer flock is the only reliable way to bring egg prices down to historic levels and have them stay there. However, this would be a massively complex undertaking that could take years to come to full fruition. This means that if Trump wants to be able to take credit for bringing egg prices down by the 2026 midterms, his administration needs to act now.

The current state of play

As previously discussed, there are a few reasons we should do a vaccination campaign for our layer flock. Firstly, it can avoid the needless killing of tens of millions birds per year, often via inhumane methods.

Secondly, vaccination is the only way to get eggs back on the shelf and ensure long-term price stability. If you'll recall, this isn't the first time egg prices have been in the news because of bird flu. In late 2022 and early 2023, prices were nearly as high as they are now. And the last time bird flu appeared in 2015-2016, there was a smaller, but still noticeable price increase. Each time prices came back down, but the underlying problem remained unsolved. Many experts now believe the virus has become endemic, with year-round detections in wild birds suggesting this pattern of price spikes will continue indefinitely we achieve herd immunity through vaccination.

Lastly, vaccination can help minimize the chance that H5N1 causes a human pandemic. Prediction markets currently put the risk of H5N1 causing a COVID-level pandemic at 4-5%. This may not seem high, but considering the multi-trillion dollar effect that COVID had on our economy, and the generally crappy time everyone had during lockdowns, it’s worth bringing down as much as possible. The Institute for Progress estimates the cost of this outcome in expected value terms as $640 billion. Vaccinating layers wouldn’t eliminate this risk, but it would meaningfully lower it.

The reason that we aren’t currently vaccinating is that it could have potentially harmful effects on international trade for chicken meat. Poultry vaccines reduce symptoms and virus shedding, but they don’t completely prevent infection, meaning that infected but vaccinated birds could still carry and small amounts of virus into an importing country. I’ve previously argued that in theory vaccinating layers shouldn’t affect broiler exports, because they’re two completely different supply chains. But alas, that does seem to be the situation.

Since I previously wrote about this issue, the conversation around bird flu vaccination has progressed rapidly. The dairy, turkey, and egg industries sent a joint letter to the new Agriculture Secretary Brooke Rollins advocating for a vaccination campaign. And the chicken meat industry sent a similar letter, requesting that any vaccination campaign come with a plan for protecting broiler exports. Brooke Rollins has signaled that bird flu is going to be one of her top priorities now that she’s been confirmed.

It’s good that these conversations are happening now because the decision of whether to vaccinate is, in some sense, the easy part. Actually getting vaccines to the over 300 million layers is going to be a massive undertaking and it will take some time for results to show up.

The logistics of mass vaccination

The Zoetis vaccine that recently got conditional approval is a H5N2 subtype1 killed virus vaccine. This means that non-infectious bits of the virus are injected into the healthy bird, teaching the bird's immune system what to look for. These types of vaccines are typically administered to a bird that’s a few weeks old, and for layers will likely have to be given in two doses for full effectiveness.

There are a couple different ways of administering vaccines to birds at large scales. From easiest to hardest, they can be administered at the hatchery, the rearing farm (where chicks grow up until they’re old enough to lay eggs), or the egg farm.

At the hatchery, vaccines can be administered in-ovo (injected straight into the embryo while it’s still in the egg) or right after the chicks hatch. In-ovo vaccination is not yet common for layers, because half of the eggs are male, and don’t need the vaccine (however, as in-ovo sexing becomes more common, this may become a possibility). More likely is that vaccines will be administered to female chicks immediately after they hatch. Layer hatcheries generally already have the infrastructure to do this, since they administer a number of other vaccines for diseases like Marek’s. The hatchery is the easiest environment in which to vaccinate because chicks are relatively easy to work with and are already being handled and processed for other things like beak treatment. However, only a few vaccines can be administered at the hatchery to prevent overloading of the chick’s immune system, so adding an additional bird flu vaccine may have some tradeoffs.

The second option is vaccination during the rearing phase, when young birds are growing but haven't yet started laying eggs. While this process requires each bird to be caught and handled individually - a substantial undertaking given the size of most flocks - rearing farms are already equipped for this task. These operations routinely administer various vaccines and have established labor protocols, infrastructure, and expertise for handling birds at this stage. Additionally, rearing farms are less densely populated and birds are often reared on the ground, making the task more feasible than vaccination during the laying phase. By default, the Zoetis vaccine would likely be administered here when birds are a few weeks old, with a potential booster dose following shortly after.

The third and most challenging option is vaccination on laying farms, where mature hens are actively producing eggs. Unlike rearing farms, layer operations rarely administer vaccines to mature birds and typically lack the established infrastructure to do so effectively. The difficulties are compounded by several factors: birds are densely packed in cages or aviaries making them harder to handle, and the vaccination process disrupts egg production. The benefits are also reduced at this stage - since immunity takes several weeks to develop and hens are already part way through their productive lives, they'll have limited time to produce eggs with protection before retirement.

Despite these challenges, producers may still choose to vaccinate mature flocks in emergency situations, particularly when faced with nearby outbreaks that put their flocks at immediate risk. This approach has been used in other countries like Mexico, even with its drawbacks.

A lot still remains to be seen about how a potential vaccination campaign would actually happen. There are a number of other vaccine candidates under consideration, some of which could be administered at the hatchery. Additionally, some vaccine candidates work via a different mechanism that would allow them to be administered through the air or drinking water, which is logistically much easier. However, we don’t currently know which of these candidates will be approved by the USDA.

If the industry primarily uses the Zoetis vaccine administered during rearing, achieving full flock immunity will take considerable time. Since laying hens have a two-year productive lifespan, if we only vaccinate new birds entering the system, it would take a full two years for all currently producing hens to be replaced by vaccinated birds.

This could potentially be sped up by vaccinating mature birds, which would be easier if the USDA approved other more logistically simple vaccine candidates. However, even in these cases, vaccines would take time to scale up, manufacture, and distribute. Therefore, it may take at least a year to see the full effects of a vaccination campaign, likely longer.

The politics of vaccination

President Trump took office less than a month ago, but if he wants to solve this issue by the 2026 midterms, he needs to act now. Otherwise, egg prices could come down temporarily, then spike again at a politically inexpedient time, at which point he’ll no longer be able to blame it on Biden.

The international trade barriers remain a serious challenge, but Trump's proven willingness to use tariffs as leverage with trade partners could help resolve the vaccination impasse. By applying pressure on countries that might restrict chicken imports over vaccination concerns, he could help ensure broiler exports don't block necessary protection for layers and turkeys. This approach might even encourage other countries to start their own vaccination programs, breaking the current prisoner's dilemma where everyone waits for someone else to move first, and be a positive step for the entire global poultry industry.

He may face pushback from certain parts of his party that seem to be allergic to the word "vaccine." However, when grocery store shelves are empty and egg prices spike, voters of all political dispositions take notice. We therefore have a rare issue where rational policy making and political expediency are aligned. Vaccination simultaneously addresses food prices, animal welfare, and pandemic risk, turning a persistent crisis into a clear political win.