It’s Summer of 2025. I’m standing in a grass covered field on the longest day of the year. A friend of mine walks towards me, holding his newborn son.

“Hey, I don’t know if you’re aware of this, but you were pretty instrumental in this kid existing. We read your blog post on polygenic embryo screening back in 2023 and decided to go through IVF to have him as a result.”

He hesitates for a moment, then asks “Do you want to hold him?” I nod.

As I cradle this child in my arms, I look down at his face. It feels surreal to think I played a part in him being here. It's the first time I've met one of these children that I've worked so hard to bring into existence.

My mind wanders back to a summer five years before when I was stuck at home during COVID, working my boring tech job selling chip design software for a large company. I remember the feeling of awe I had upon learning that it was possible to read an embryo’s genome and estimate its risk of conditions like diabetes, then choose to implant an embryo with a lower risk.

I remember the struggle of trying to break into this new field of reproductive genetics, one I knew nothing about. I remember the endless hours reading research papers, doing computational modeling, the meetings, the flights, and the endless debates about the ethics of trying to bring children like this into the world.

As I stand there in the shade of a large oak tree, the wind gently blowing past us, it all feels oddly distant. I watch his tiny little fists clench and unclench as his eyes drift back and forth gazing at the branches above us.

“You have no idea how loved you are little man”, I think to myself. “Or how many people that you will never meet worked to make it possible for you to be here.”

I hope some day he will understand. He’s among the first of a generation of kids that will have a chance to grow up a little healthier, a little smarter, and hopefully a little happier than the children that came before him. Maybe someday most children will be like him.

In just the last two years, I’ve watched polygenic embryo screening, the technology used to make this possible, explode in popularity among my social circles. A good third of my friends are now doing IVF just to get access. Almost every month there's a record number of new parents signing up.

Everyone has a different reason for wanting it. Some have a family history of some disease and they really don’t want their kid to get it. Some want to make their kids smarter. Some really want a girl.

By now, I’ve chatted with dozens of parents going through this process. They gather together at dinner parties in San Francisco, or in Signal group chats and excitedly go over the scores for their latest round of embryos. They trade tips for how to get medications cheaply, which IVF doctors to go to, and what supplements to take.

There are a million little things to learn. Some IVF clinics can get you twice as many embryos as others. Some are a third of the price. Some will literally fire you as a patient if you tell them you are doing embryo screening.

And then of course there are the genetic testing companies themselves, which have major differences. One of the companies has significantly better genetic predictors. One is much cheaper. And one spent four years misleading its customers about their genetic risk, and trying to cover it up.

Over the past half decade, I’ve worked at two of these companies, started and ran an embryo gene editing company, co-founded a company that helps parents find the best IVF clinic, and helped dozens of parents go through embryo screening. I’ve learned quite a bit about this technology, what it can and can’t do, and how to optimize the process. I’m going to tell you what I’ve learned along the way.

How large are the benefits of embryo screening? Is it even worth going through IVF?

Embryo screening (also known as PGT-P), can increase expected IQ by about 4-10 points, increase life expectancy by about 1-4 years, and decrease the risk of many diseases by about 10-85%. It can also make your kids taller, reduce their risk of mental health disorders like autism or schizophrenia, and (probably within the next year) make them slightly more extroverted or less neurotic.

Exactly how large are the benefits? It mostly depends on two factors: how many embryos you have and the strength of the genetic predictors you’re using to select one.

The strength of the genetic predictors mostly depends on which embryo screening company you use and your genetic ancestry. The number of embryos you get depends on how old you are, how many rounds of IVF you go through, and how well you respond to stimulation protocols.

But these are just words. To help you better estimate the benefits given your own situation, I’ve embedded a calculator below. Keep in mind this is showing expected benefit, and that in your specific cycle the benefits could be higher or lower.

Most parents care about more than one trait. Many will care about having a smart kid, but almost everyone cares about mental and physical health. So how does this work in practice?

The answer is you get a report showing how each of your embryos scores across multiple dimensions, and you choose which tradeoffs to make. You can see what that looks like by clicking the "Display all traits" button in the calculator above.

An actual report from one of the companies will look something like this:

Embryo 9, for example, has a slightly increased IQ, a slightly elevated risk of ADHD, an average risk of bipolar disorder, and a very high risk of schizophrenia.

How do parents decide whether a few extra IQ points are worth the increased risk of those other disorders?

In this case the choice is actually pretty clear; a 9% lifetime risk of schizophrenia is probably not worth an extra point or two of IQ, so almost all parents are going to deprioritize this embryo for transfer.

But in other cases it’s not actually clear which disease is worse or how to think about tradeoffs between them. How exactly do parents deal with this? In practice, there are a couple of methods.

The nerds use spreadsheets. Anfei Larkin has written about how she and her husband went through every trait and disease the company offered and tried to reason through things like what percent extra lifetime risk for Alzheimer’s balances a one percent decrease in lifetime risk for bipolar disorder.

In the end, they averaged their weights for each trait and plugged all the numbers into a spreadsheet to produce a final ranked list of the embryos.

Most parents are not quite this meticulous. The majority of those I’ve spoken with do something like rule out embryos that have exceptionally high risk of serious diseases, then pick the remaining one that scores best on one or two metrics they care about most (usually IQ or a specific disease that runs in their family).

One of the interesting things about embryo screening is that average risk reduction doesn't always tell you the full story. Sometimes the risk reduction can be substantially larger or smaller than the numbers might suggest. I’ll tell you an interesting story to illustrate why.

When averages don’t work

Marshall and Victoria Fritz are a couple I met through Herasight, one of the embryo screening companies working in this space. Victoria has type 1 diabetes, a condition she is hoping to not pass down to their children. Genetics play a major role in T1D, and most people that have the disease possess two high risk variants in their HLA gene, an important part of the immune system.

Victoria is no exception. This would usually mean her children would be at elevated risk of the disease. Embryo screening is already a great tool for reducing T1D risk. We understand the genetics of the disease better than almost any other polygenic trait.

But for Victoria, embryo screening wasn’t just good. It was incredible.

You see, Victoria happened to marry the perfect guy. Her husband, Marshall, has the single most protective T1D genetic variant that we know of. One that reduces the risk of the disease by about 97%.

As a result, they have four embryos that are almost guaranteed to avoid the disease.

This will sometimes be the case for specific diseases where single genetic variants have large impacts.

At least two companies in the space, Herasight and Orchid, offer products that allow parents to get sequenced before they commit to paying the full price for embryo screening. If you’re specifically interested in screening against a particular disease, it’s plausibly worth booking a call with one of them to figure out if the disease you’ve got in mind has this property.

Here’s an incomplete list of these kinds of diseases where knowing the genetic background of you and your spouse ahead of time can be unusually useful:

Age-related macular degeneration
Alzheimer's
Parkinson’s
Type 1 diabetes
Celiac disease
Type 2 diabetes
Inflammatory bowel disease
Psoriasis
Breast Cancer

How much does IVF cost?

For most couples with a woman in the 25-35 age range, you’ll need to go through 2-4 rounds of IVF to get enough embryos to select from. Each of these rounds will cost $14,000-$40,000, with the variance coming almost entirely from which clinic you choose to go to. You’ll then have to pay for genetic testing, which is an additional expense of about $7000 to $70,000 depending on what you want to screen for and how many embryos you have. In all, most people will pay $55,000-$150,000.

The two biggest cost drivers are your choice of IVF clinic and your choice of genetic testing provider.

IVF clinics have pretty significant variability in cost, with the lowest ones being around maybe $14,000 per cycle and the highest being around $40,000. In a typical medium to large size city, costs will look something like this:

Hat tip to Sam Celarek for calling up a dozen IVF clinics in Boston to gather this data.

These costs can often be spread out across a number of years. If you’re a single woman or you’re not certain who you want to have kids with, you can freeze your eggs for about half the cost of freezing embryos, and postpone paying for embryo creation and testing.

You might think that clinic cost reflects clinic quality. And while there’s some truth to that, it’s not strictly the case. Some IVF clinics are quite a bit more cost-efficient than others. A year ago, my friend Sam Celarek put together this graph of expected cost per child at different IVF clinics in Boston after combining data from his clinic success rate model with pricing info he gathered by calling a bunch of different places in the city.

The actual price per child for couples without infertility will be a bit lower than this (Sam built this graph using success rates from infertile women), but the data reflects something real: some IVF clinics are much more cost-effective than others.

How to find an IVF clinic

This naturally raises an important question: which IVF clinics are most cost-effective?

You can ask Sam for access to his data, but one obvious contender is CNY Fertility, a low-cost clinic with branches in Georgia, Florida, Colorado and New York. You can do a round of embryo freezing with them for about $15,000 including the cost of travel, medications, monitoring and the procedure. You can also do egg freezing for about $6500.

CNY’s overall success rates are a bit below average in Sam’s model, but this is probably somewhat confounded by the types of patients it attracts. They’re well known for taking basically anyone, including patients other clinics will outright reject. Sam has tried to factor this in by controlling for differences in patient age and a few other variables when comparing clinics. But it’s not yet possible to control for everything, and my best guess is this is dragging down their numbers a bit.

Still, you’re more likely to need to do an extra round at CNY than at a top tier clinic.

So where should you go if you want the absolute best results? What’s the best path if money is no object and you're willing to pay more per embryo to get the whole thing over with faster?

In that case, the best doctor I know of is likely Dr. Aimee, a fertility doctor in San Ramon CA who has done IVF cycles for many of my friends and seems to generate outlier results at a surprisingly frequent pace. One woman I know who went to her managed to get 17 euploid embryos from a single IVF cycle, an embryo shy of the highest number I ever saw while working at Genomic Prediction. She’s also been able to get a really crazy result for a friend of mine that had infertility issues: this friend had done 6 prior rounds of egg retrieval, getting 0-1 euploid embryos each time, until she went to Dr Aimee. She then had one round that produced a single embryo, after which Dr Aimee added a particular medication to her retrieval process. Her next two rounds produced 5 euploid embryos each time, which if you know anything about IVF outcomes, is an insane jump.

Her prices are on the higher end, but if money is less of a concern for you and you're in the bay area, she's likely worth it. I don't know her exact prices, but it seems to be between $25k and 40k per round.

None of this is a guarantee by the way. My sample size for the above statements are based on outcomes from about ten women.

What about if you’re not in the bay, and you don’t want to travel?

In that case, your best bet is picking a clinic using Baby Steps IVF. Baby Steps IVF was built slowly over the years following publication of my original guide to having polygenically screened children based on research I had done into the success rates of different IVF clinics in the US. Sam Celarek took my raggedy model published in 2023 and turned it into a proper machine learning model using much more advanced statistical techniques and additional data from both SART and the CDC.

It’s now probably the single best resource online for quantifying how good different IVF clinics are. And we’ve made the basic clinic rankings available for free.

Thanks to a lot of work from him and Roman Hauksson and a grant from Astral Codex Ten, this has now become an accessible resource for anyone going through IVF.

I’ll have more to say about Baby Steps in a future post.

Which PGT company should I use? What are the advantages of each?

I should start this section by noting that during the process of drafting this post, I joined Herasight, one of the four main polygenic embryo screening companies. I also have a small amount of stock in Genomic Prediction where I was previously an employee in 2022.

Nonetheless, I’ll do my best here to provide an objective analysis of the current state of the industry.

There are four main companies offering polygenic embryo screening, and a fifth newer one that just got started recently.

The four main players are Herasight, Orchid, Genomic Prediction and Nucleus. The fifth company is Reticular, which is hoping to apply techniques used to create interpretable AI for biology to embryo selection. They're quite new and to the best of my knowledge haven't published any validation papers, so the rest of this post won't mention them.

The shortest summary ever of the four main players is: Herasight has the best predictors and offers selection for the most things but is expensive, Orchid is good but is also fairly expensive, Genomic Prediction is not quite as good but is the cheapest by far, and Nucleus is sketchy and should probably be avoided. I've included a slightly more detailed table below and a much more detailed writeup in the rest of this section.

Quick comparison table

	Herasight	Orchid	Genomic Prediction	Nucleus
Predictor strength	Strong	Good	Good	Unclear, misleading estimates for many traits/diseases
Predictors validated within family	✓	For some diseases	For some diseases	For some diseases
Screening for non-disease traits?	✓	X	X	✓
Genotyping method	Whole genome sequencing of embryos or imputation from PGT-A data	Whole genome sequencing of embryos	SNP array	Probably SNP array
Can detect de novo mutations?	Yes*	Yes	No	No
Works with data from any genetic testing company?	Yes	No	No	No
Sequencing of families?	Yes, including extended family for premium product	No	Yes, parents only	Yes, parents only
Long-read sequencing	Yes	No	No	No

*Herasight can only detect de novo mutations if your clinic sends embryo biopsies to their lab. If you use their ImputePGTA product you won't be able to get de novo detection.

Price comparison

Notes on the above graph

Genomic Prediction uses a different genotyping method from Orchid and Herasight, so it has some additional limitations that aren't captured in this price chart. I get into the differences in the section below this.

Orchid Platinum doesn't seem to have a standard pricing yet. I've shown the cost per embryo based on the one report from a person I know who has used it, but it's possible you might be offered a different price by the company. Last I heard it was $10,000 per embryo, with a $7500 discount if the embryo comes back aneuploid.

Herasight Health is Herasight's new "health only" product that is less expensive than their standard $50k product, but doesn't include screening for IQ, height, or ADHD. It starts at $20k, which includes screening for 10 embryos, two of which they'll screen for de novo mutations. If you want to screen additional it's $1500 per embryo, and if you want to screen additional embryos for de novos the price is $2500/embryo (on top of the $1500 base fee).

Herasight also has a standard product but it’s not shown in the chart because it doesn’t vary in price with the number of embryos. The standard price is $50k and it includes everything the company offers (IQ, height, all diseases, expanded carrier screening and universal PGT-M, family sequencing, de novo mutations, polygenic scoring of parents, access to the embryo simulator, updates for the next five years, and a few other things).

What are the actual differences between the embryo selection companies?

There are fairly large differences between the embryo selection companies, only some of which are publicly known. The differences mostly boil down to predictor accuracy, cost, and the set of things which the different companies screen for.

Every capability offered by the different companies is downstream of their ability to read an embryo's genome accurately. And each of the three main companies has a fascinatingly different approach to this problem.

How Genomic Prediction reads a genome

Genomic Prediction uses SNP arrays, an older technology that looks specifically at spots in the genome that commonly differ between people. They measure about 1.7 million locations in the genome to detect which of two variants a given embryo has.

Given the human genome is 6 billion base pairs, you might be surprised that this technique works at all. How is it that we can figure out someone's risk of diabetes or schizophrenia when we're only looking at 0.03% of their genome?

There are two answers to this problem:

First, any given human genome only differs from another along just 0.1-0.5% of its length (0.1% if you only count single base pair changes, 0.5% if you include structural variants such as regions where one person might have an extra copy of a gene).

Second, it's possible to do a decent job inferring the value of genetic variants you don't directly read by making educated guesses about them based on the portion you do observe.

Due to quirks with how sperm and eggs are formed, you can use the measurements of an embryo's genome you do have to guess at the part of it you don't have.

Your ability to make these guesses well hinges on the quality of the reference panel, and since these reference panels are full of mostly European genomes, our ability to guess at the rest of the genome is not as good for non-European individuals.

SNP arrays are also fundamentally limited in their ability to detect rare variants, structural variants, and highly diverse parts of the human genome like the HLA region I mentioned in the story about type 1 diabetes. So there is in fact a reasonably significant tradeoff to using them. The size of this tradeoff has actually grown over time because we've recently gotten a better understanding of the role of rare variants in predicting life outcomes, and Genomic Prediction can't really take advantage of those modern advances. But despite all these limitations, Genomic Prediction can still do an OK job predicting disease risk from a genome.

How Orchid reads a genome

In the early 2020s, Orchid Health became the first competitor to Genomic Prediction in the polygenic embryo screening space, launching a significantly more expensive test whose main differentiating factor was offering whole-genome sequencing of embryos.

Getting a whole genome sequence of an embryo is not a trivial task. To read an embryo's genome, you need to remove 3-5 cells from an embryo around the 5 day mark, at which point the entire embryo is only about 100 cells.

High quality DNA sequencing requires reading the same stretch of DNA 30 or more times. And the DNA is destroyed in the process.

How do you read DNA 30 times when you only have 3 to 5 cells? The cells literally don't have enough DNA for you to read it 30 times.

The answer is that we must first make copies. This step is called amplification. We cut open the cells with enzymes and throw out everything except the DNA. Then we make copies of the DNA until we have enough to work with.

But this process is tricky. The amplification process introduces errors into the sequencing process. Sometimes it amplifies one region of DNA 100 times and another part not at all. Sometimes it makes errors when copying, and copies get copied and you end up thinking there's a mutation in the DNA even though there wasn't.

Orchid (and Herasight for that matter) handle this better than most: they both utilize techniques to reduce these issues with amplification. But it's still not completely perfect, and occasionally Orchid will not be able to resolve a variant because they only read its value a few times. Herasight doesn't suffer from this problem due to their heavy use of long-read sequencing, which I’ll explain in more depth later.

Whole genome sequencing allows us to better predict diseases and traits in embryos because we don't have to guess at the value of variants not captured by SNP arrays. We can directly measure them.

This makes the performance ceiling for whole genome sequencing fundamentally higher than the performance ceiling for SNP arrays. And as our understanding of the role of rare variants in disease risk and trait values has continued to improve, this gap in performance between whole genome sequenced embryos and SNP arrays has grown.

How Herasight reads a genome

Herasight has perhaps the most interesting method to figure out what's in an embryo's genome. Like Orchid, Herasight can do whole genome sequencing in their own lab. But they have an additional, much more convenient method that works on data generated by almost any genetic testing lab, including those who don't offer polygenic embryo screening.

Early in its history, Herasight was working exclusively with embryo data generated by other companies like Orchid and Genomic Prediction. They would take this data, and run an after-market analysis on it, utilizing their stronger, broader set of predictors to better choose embryos with low disease risk, and to screen for things that other companies didn't offer (such as IQ).

However, most parents doing IVF didn't have embryo genomes generated by either of these companies. If they had genetic data at all, it was the more basic type: ultra-low coverage PGT-A data used mainly to detect down syndrome and other chromosomal abnormalities.

If genetic sequencing produced a painting, Orchid would produce a Da Vinci, Genomic Prediction would produce a sketch by a promising young artist, and PGT-A would produce a crayon drawing of a cow.

For many years, these customers were simply out of luck; if you already had frozen embryos, it was impossible to test them again without hurting their chances of becoming a baby.

Sometime in 2024, Michael, the CEO of Herasight, had an interesting idea. Perhaps if he had a high quality genome of the parents, he could use that data to fill in the missing gaps in the embryo genome.

If this could be done, the implications would be enormous. Virtually overnight, hundreds of thousands of parents with frozen, PGT-A tested embryos would suddenly be able to test them for risk of almost every major disease and many of the most impactful traits; everything from diabetes to Alzheimer's to autism to intelligence could be estimated in existing embryos.

Every genetic testing company Michael spoke with about this idea either didn't understand it, or thought it was impossible.

But the company persisted, and in 2025 they managed to get it working. The resulting ImputePGTA algorithm unlocked polygenic testing for hundreds of thousands of couples with frozen embryos in storage.

ImputePGTA is a major breakthrough for parents who already have frozen embryos, but there are also significant benefits for parents who are planning to do IVF soon. These parents no longer have to go to a clinic that works with Genomic Prediction or Orchid (or even Herasight). So long as a clinic does PGT-A and is willing to transfer an embryo chosen by the patient, you can do polygenic embryo screening. This is actually a pretty huge deal, and has opened up PGT-P to customers all over the world.

There’s a special type of sequencing that must be done to make ImputePGTA work, and it’s one that no other polygenic screening company offers; long-read, high depth sequencing of the parents.

Long-read sequencing, especially the high depth kind that Herasight does, is among the most expensive ways to read a genome. The sample preparation method takes days, there are expensive consumables involved. And if you mess up any part of the process you have to do the whole thing over again.

It’s a pain in the ass. But it allows them to do something no other company can do: it allows them to fully phase a parental genome.

Phasing is when you not only read which genetic variants are present in a person's genome, but you identify which of the two chromosomes a given variant is part of. This is extremely important for ImputePGTA because you’re inferring which part of each parental chromosome an embryo got based on crappy PGT-A data. That’s only possible if you know which variants are paired together on the same chromosome in the parent.

But long reads also have some other benefits that aren't widely understood. Some parts of the genome are extremely repetitive for long stretches, and whether or not you get a disease hinges on how many times a given sequence repeats.

If you're reading stretches of DNA 150 bases at a time, you often can't tell how long the repeat is. This is especially important for conditions like Huntington's disease, fragile X syndrome, and spinal muscular atrophy. Herasight is uniquely good at reading these sections of the genome, allowing them to diagnose essentially all codon repeat disorders. This is one of the small ways in which they have an advantage over Orchid specifically when it comes to rare disease.

A full list of such diseases is beyond the scope of this post, but if you have a specific interest in one of these conditions you can just book a call with them and they’ll put you on the phone with someone who can walk you through what’s possible.

Genetic load testing, de novo mutations, and other differences between embryo screening companies

Every new generation of children receives ~100 new genetic mutations that their parents didn’t have. These genetic anomalies are referred to as “de novo” mutations, because they aren’t really transmitted to the child via typical inheritance. Instead, they come from a DNA copying error or (less commonly) from a cosmic ray hitting your dad in his balls.

Most of these mutations will have approximately no effect. Usually they only change a single letter in a part of the genome that isn’t directly translated into a protein, thus the effect will be small or zero.

But not always. Every now and then one of these mutations will hit a protein coding region and change the sequence of amino acids. And sometimes these mutations will occur in very, very important genes.

This still doesn’t necessarily cause problems. Sometimes you'll get lucky and the mutation will be silent, meaning the protein won't be changed.

But sometimes you don't get lucky, and the results can be catastrophic.

Among these very important genes are a set of proteins called "highly conserved sequences". They’re called that because they're virtually identical across most if not all humans, and if their biological function is fundamental enough, most mammals. Sometimes they're even identical across whole branches of the tree of life.

Mutations in these genes can lead to anything from a mild physical or mental impairment to a developmental disorder so catastrophically bad it literally ends a pregnancy before birth. And it can lead to anything in between.

Fortunately for the human species, these kinds of mutations are rare. Our best guess is only about 0.34% of children will be born with such a mutation. The rate is significantly affected by parental age, especially the age of the father, who is the dominant source of these mutations.

Reducing the chance of these sorts of issues (along with remaining more fertile for longer) is one of the strongest arguments for freezing your sperm. If you freeze your sperm around age 20 (or better yet sometime in your teens), you can cut the number of de novo mutations your children get by a third or more.

Orchid was the first company to offer screening of de novo mutations several years ago. Herasight followed suit recently, though it should be noted that this is not available if you use their ImputePGTA product (they need to be able to sequence the embryos themselves to call de novos).

I am probably somewhat biased here, but I think Herasight’s tech to classify the pathogenicity of de novo mutations is probably the best currently on the market. They recently released a predictor called “NeuroRisk” which to the best of my knowledge is better than any other predictor to date at identifying mutations that cause severe developmental disorders, including non-verbal autism.

Family history

Herasight has some very neat tech that takes your family history into account when predicting your embryo's risk of a disease. They can look at family members who have been diagnosed with the condition and figure out how much DNA they have in common with each embryo. This allows them to estimate disease risk even when they don't know which DNA changes are causing the change. So far as I know, they're the only ones doing this.

Herasight also takes family history into account when estimating the baseline risk that one of your embryos gets a disease. When I last checked about a month prior to this post, Orchid wasn't doing this. In practical terms, this means Orchid is more likely to underestimate an embryo's risk of a disease for which you have a family history.

Expanded carrier screening and universal PGT-M

About 15% of couples are carriers for a high impact single gene variant. Conventional genetic testing basically never catches this stuff unless you already know you're a carrier and order a test specifically for a known issue.

For the rest of us, it's a crapshoot. You conceive naturally or via IVF, and if you or your spouse are in the 15% that carry one of these things, each of your kids will have a 50% chance of getting it from you.

These mutations vary in their significance; sometimes they're relatively benign (you'll have a significantly increased risk of some condition in old age), and sometimes they're very, very scary (sudden death from heart failure in your 40s, dramatically increased risk of non-verbal autism). A few of these are well known and well understood; the gene that causes Huntington's disease, for example, has been understood for decades.

Most of these are dominant variants, meaning they’ll actually manifest in the parents. This can actually be a life-changing diagnosis: Herasight has, in more than one case, identified a pathogenic variant that causes massively increased risk of sudden heart attack in their customers. In several cases, the conditions whose risk was increased by these variants had preventative treatments that could substantially reduce risk.

Without a test like this, it's very common for neither of the parents to know there's a potential issue until they get sick. And even when they do become symptomatic, the doctors don't always identify the issue correctly. And even when they do, undoing the damage can be difficult or impossible.

We've been able to test for some of these variants for decades, but only if the IVF clinic knew to look for it (which usually meant a known family history for a specific condition and confirmation that one of the parents was either a carrier or affected).

The lack of universal screening for high impact monogenic variants is one of the most tragic failures of modern IVF. Extrapolating from 2019 data and the recent growth rates of IVF, there were probably around 200 children born per day with a serious genetic variant that likely could have been avoided.

This article is mainly about polygenic embryo screening and all the amazing things you can do with it, but if you’ve got one of these high impact monogenic variants, that all goes out the window until you’re certain your child will not carry it.

For all their importance, we know shockingly little about these variants. I wish I could show you a graph of the distribution of the impact of these sorts of things, but we're just now starting to understand this stuff. I don’t want to brag too much here (and I can’t, since I didn’t really do any of the work to make this happen), but Herasight probably has the best tech for screening these kinds of variants at the moment. If anyone is interested in learning more about this, reach out to me via email or better yet just book a call with us using our contact form.

Orchid also screens for this kind of stuff, but I believe at the moment their panel (at least for the main product) is more limited than the one offered by Herasight. Orchid Platinum offers screening for a broader set of genetic variants, but from what I’ve heard it can be very, very expensive ($10,000 per embryo last I heard).

What’s the deal with Nucleus?

You might have noticed I’ve been avoiding mentioning a certain company up until now: Nucleus Genomics, probably the most well-known company currently offering polygenic embryo screening.

The reason for this is actually pretty straightforward; Nucleus has almost certainly been misleading customers about how good their genetic predictors work since they were founded in 2021. The CEO has, as one might expect, denied all of these allegations. But it would be pretty straightforward for them to release convincing evidence that their predictors actually work, and they haven’t done it.

To give a simple example, here’s a graph showing the performance of Nucleus’s genetic predictors implied by their reports vs the actual performance when their predictors are assessed:

The fact that the red bars are lower than the blue bars for all but one condition (schizophrenia) indicates that Nucleus is systematically misrepresenting how well they can predict genetic risk in their customers.

This is obviously a major problem. If your company’s entire value proposition is that you can predict an adult or embryo’s risk of a disease, but you report is miscalibrated on almost every single condition, that’s really bad!

Nucleus has taken SOME steps to improve things since they launched their embryo screening product. They released the Nucleus Origin white paper, the company’s first actual validation of their genetic predictors. The paper does within-family validation of genetic predictors for several diseases, it produces strong results, and it even quantifies the general level of drop-off in performance for non-European ancestry groups (though it doesn’t give a condition-by-condition performance comparison for these groups).

But there are still problems. For one thing, they’re not just offering embryo screening for the diseases validated in their white paper. If you go to their website, you’ll see they offer screening for tons of things. And it would not at all be surprising to me to hear that those predictors probably suffer from the same miscalibration as the ones they were using previously.

Second, Nucleus appears to still not have rolled out their updated predictors for adult testing, meaning they’re still using the crappy low quality predictors on paying customers even after publishing the Origin white paper. It has been five months since its publication. I don’t understand how they have moved this slowly.

Maybe at some point Nucleus will fix their problems and offer a good product. But despite having raised $32 million, the CEO has so far proven either uninterested or incapable of doing so. I’m not holding my breath for that to change.

How do I do this? Where do I start?

If you want to do polygenic embryo screening, and you’re in a position where you can handle the financial and physical costs, the best place to start is probably by scheduling a consultation with one of the polygenic embryo screening companies.

Get in contact with Herasight here
Get in contact with Orchid here
Get in contact with Genomic Prediction here

Any one of those companies can give you a basic overview of the process and tell you which clinics you can work with to do polygenic embryo screening, or how to ask a clinic to send biopsies to their lab.

To answer the all-important question of “will this clinic transfer an arbitrary embryo of my choosing”, you’ll usually have to book a paid consultation which typically runs between $250 and $500 in the US. If you get answers for a specific clinic, please let the rest of the world know the results! I’d like to put together a list of different clinics that support PGT-P, including info about which PGT providers they’ll work with. Any info you can provide to the rest of that helps support other people going through embryo screening.

Once you’ve found a clinic you want to use that will let you transfer an embryo of your choosing, and you’ve gone through the first consultation, you’ll begin the testing process.

One of the very first things a clinic will do is take a blood test and do an ultrasound. Two of the most important measurements they’ll take here are your AMH and your AFC. Both of these measurements can give you a rough idea of how well egg or embryo freezing is likely to go for you. And generally speaking, higher is better.

Assuming all looks good, you’ll start hormone injections to prepare for egg retrieval usually within a month or two of the initial consultation. After 10-12 days of injections, ultrasounds, and blood tests, you’ll travel to the clinic to get your trigger shot. This shot usually contains a mixture of lutenizing hormone (LH) and human chorionic gonadotropin (hCG), and allows the egg to be released from the ovarian follicle in which it’s contained.

About 36 hours later, you’ll return to the clinic for the actual retrieval, during which a doctor will retrieve eggs from your ovaries under the guidance of ultrasound. They’ll usually use local anesthetic for this procedure, meaning you’ll usually be able to return to your work or normal life within a day of the procedure.

There’s a surprising amount of variance in the side effects people experience from egg or embryo freezing. One woman I know had basically zero symptoms from multiple rounds of egg retrieval. Another found herself doubled over on the bathroom floor in pain during one particularly bad retrieval.
Most women experience something in the middle: mild bloating and abdominal discomfort along with some degree of mood swings. This usually is on the level of “bad period”.

The two worst side-effects of egg retrieval are internal bleeding and ovarian hyperstimulation syndrome. Internal bleeding of the type that requires medical intervention is pretty rare; less than 0.1% of cycles. Ovarian hyperstimulation syndrome is more of a continuum, with mild bloating on one end and hospitalization on the other. Hospitalization from egg retrieval is fairly rare; about 1%. There are also strategies for preventing the more severe forms of OHSS, such as taking letrozole during stimulation if your e2 levels get too high, and taking cabergoline after retrieval to reduce bloating.

You’ll likely want to do 2-4 rounds of retrievals to create all the embryos, with the exact number varying fairly substantially depending on how well you respond to stimulation and how many children you want. I’ve had friends do one retrieval and I’ve had friends do ten.

How to get cheap IVF medication

IVF medication usually runs between $4500 and $6500 per round. If this sounds expensive, that’s because it is. It usually makes up 20-30% of the cost of IVF. But it’s an area where an enterprising woman can significantly bring down costs.

Believe it or not, one of the best places to currently get major discounts on fertility medications is TrumpRX. The website launched in early February of 2026, and as of this writing, it has significant discounts on several major IVF medications like Gonal-F, Cetrotide, and Ovidrel.

GoodRX also often has discounts on drugs like Ganirelix, Leuprolide, and Endometrin.

Specialty pharmacies often have very good deals on particular medications. IVFPharmacy.com has very good pricing on Follostim. According to some old forum posts, Mdrusa.com apparently had very competitive pricing on Menopur. If you want to check their current prices you'll need a prescription.

If you put all these together, you can just about halve the price of IVF medications.

There's one last source of cheap meds I haven't mentioned yet. It's cheaper than all the rest, sometimes by a factor of 20: Peptide chat.

Peptide chat is a gray market supplier of “research chemicals” including the “Chinese peptides” everyone has been going crazy for lately. One customer I know bought Menopur from them for 95% below the listing price on the cheapest conventional pharmacy. She had several IVF cycles with stupidly good results while taking these meds (though this is very likely just a result of her unusual biology rather than because the medications they supply are extra special). She got Menopur from Semag Peptide (it’s labeled HMG and sells for $70 per 750 IU).

I know a few people who have used Peptide chat for various medications, so I know at the very least that they’re not usually lethal or a scam. But this is still very much a "use at your own risk" type of medication supplier, so don't take this as an endorsement of them from me. It's simply a possible source of medication you may want to look into more.

All together, medications will run between $300 at the very low end and $6500 at the very upper end. $2000 or so is about the lower limit for traditional sources of medication. If you want to go lower than that you’ll need to get meds from the gray market peptide dealers or from women selling their leftover meds on Facebook.

Connecting with me and others in this process

If you’ve got questions about IVF, feel free to either leave a comment or send me an email at genesmithlesswrong@gmail.com. I know quite a few women doing egg freezing or IVF who are interested in polygenic embryo screening, so if you’d like to connect with others going through a similar experience, let me know and I may be able to add you to some group chats.

If you’re in the bay area and would like to meet up in person, feel free to drop me an email. I’m always happy to chat with people who are serious about doing polygenic embryo screening, or who are interested in working in the field.

FAQ

Is this post medical advice?

No

Are IVF babies less healthy than naturally conceived babies?

The short answer is probably not, it's hard to rule out entirely because there are no proper randomized control trials. There are many studies showing the average child born via IVF has more problems, but almost every single one I've looked at is confounded in some pretty obvious way.

Historically, parents who use IVF are older. They're sicker. They often suffer from repeated miscarriage or have uterine issues which can lead to premature birth. And the (increasingly uncommon) practice of transferring multiple embryos at a time led to high rates of twin pregnancies from IVF, which themselves are known to carry additional risks for the babies and the mother.

I’ve looked at about half a dozen studies on this topic, and essentially all of the ones that find harm don’t properly control for these kinds of selection effects.

There’s one exception, which is a Nordic study showing a higher rate of childhood cancer when using frozen embryos rather than fresh. I couldn't fully explain this effect with selection effects, though there may be something I'm missing.

The absolute risk increase was very small; I believe around 0.2%. And there are some other reasons why this result might not replicate in a modern setting: the freezing protocols used in most of these cycles were the old, slow freezing methods rather than modern vitrification techniques to freeze the embryos.

Still, this is the one result I've looked at where the authors seem to have done a decent job adjusting for confounders, and they still found this a negative effect. So although there's not much evidence that IVF itself is bad, we can't rule out negative effects completely.

This whole topic deserves further elaboration, but I’ll have to save that for another post.

How do we know embryo selection actually works?

More skeptical readers might wonder how we know embryo selection actually works. How can we actually validate that these genetic predictors work if the first polygenically screened baby was only born in 2019?

The answer is we test the predictors in existing people. So we get a bunch of old people from UK Biobank or some other source of data, make predictions about which of them have cancer, then check to see how well we did.

"But", you might contend, "you're just seeing that the genes you've identified are correlated with disease. How can we be certain they're actually causing that observed increase?"

This is a very good question, and in most fields we wouldn't be able to go any further than this. Except in genetics, we actually can test whether they're causal. To do this, we test our predictor in siblings.

Siblings have this amazing property, which is that they inherit a randomized subset of DNA from each of their parents. Importantly, the subset they get is not affected by the subset their sibling got. So if we test a diabetes predictor and we can predict which sibling has diabetes better than random chance, we can be sure that we truly are picking up a causal signal.

This is why sibling validation is so important, and why you should generally trust companies that have extensive sibling validation more than ones that don’t.

If I want to use a cheaper clinic, do I need to spend 3 weeks traveling?

No. Most clinics (especially ones that see many travelers) offer something called “remote monitoring” which allows you to do the required ultrasounds and blood tests at a facility close to your home, and only come to their clinic for the last 3 days or so of your retrieval. This is something worth asking about during your consultation.

Which clinics definitely offer polygenic embryo screening?

Every polygenic embryo screening company has their own list of clinics they work with, but they're cagey about sharing it publicly. So unfortunately if you want to figure out whether a nearby clinic will work with them (specifically by sending embryo biopsies to their lab), you'll have to either call the clinic or the screening company directly. Generally speaking it's better to contact the screening company, since not everyone at the clinic will know which PGT labs they work with.

Which clinics definitely DON'T offer polygenic embryo screening?

Spring Fertility in San Francisco once fired a friend of mine as a patient after she mentioned she wanted to do polygenic embryo screening. I would strongly recommend not using them for this process.

How many rounds of IVF should I do?

Generally speaking, you want at least 2x as many euploid embryos as you want children. Any fewer than this and you’re not going to get much of a benefit from polygenic embryo screening.

What should I do with my immature eggs?

Freeze them. Most clinics will throw them out. But there’s good technology to mature between 40 and 70% of your immature eggs and it’s worth using!

In most IVF cycles, about 10-30% of the eggs harvested will be immature. Normally these just get thrown out. Sometimes if a patient is really desperate, or if they’re doing a special protocol called “low-stim”, the clinic will save these eggs and try to mature them with some special liquids.

This doesn’t usually work that well. Traditional IVM liquids can only turn about 40% of immature eggs into mature ones. Even when they succeed, the resulting eggs develop into implantable embryos at a lower rate.

A year ago a company called Gameto announced that they had a new way to turn immature eggs into implantable embryos at about twice the rate that was possible with previous techniques. In May of 2025 they announced the birth of the first baby created from an egg that was matured using Fertilo, as part of a clinical trial.

As of this writing, Fertilo is available in Mexico, Peru, and Australia. It likely won’t be available in the US until 2027, at which point you’ll be able to get it prescribed off-label by your doctor. If you save up immature eggs from multiple cycles, there will likely be a way to dump all those immature eggs into a Fertilo bath and mature about 70% of them.

This will probably boost the number of eggs you get per cycle by 10-20%. It’s not a huge difference, but clinics don’t tend to charge you more for freezing immature eggs, so you might as well do it.

Should I save some eggs or should I just make embryos?

I think there’s a reasonable chance we can get embryo editing or sperm screening working in the next five years. Embryo editing would allow you to get rid of genes like APOE e4 for alzheimer’s or some BRCA breast cancer variants, or some rare diseases, but it will only work on freshly fertilized eggs. If you’ve already got frozen embryos you probably won’t be able to utilize editing tech.

Sperm selection will involve selecting the best sperm and fertilizing an egg with it. This tech is obviously still speculative (no one has gotten anything working at the moment, but if it did work it could potentially double the expected gain from embryo selection). You won’t be able to use sperm selection if the eggs have already been fertilized.

For this reason it may be worth freezing some number of extra eggs just in case one or both of these technologies come online.

Can I do embryo screening outside the US?

Yes, you can do embryo screening outside the US. I know several couples that have done it in Europe, and a few elsewhere, such as Dubai.

There are some countries where it’s actually just impossible: Germany being perhaps the most notable. The UK HFEA has also published guidance stating that polygenic embryo screening is illegal, though it’s not entirely clear whether they really have the statutory authority to do so. Australia is similarly prohibitive of this tech at the moment.

Polygenic embryo screening is feasible in most of the rest of Europe, though the degree of feasibility varies a bit by country.

Which supplements should I take?

It's probably worth talking with your doctor about CoQ10 and NAD+. A lot of the other stuff doesn’t have much evidence behind it, though I’m increasingly thinking that Rapamycin has potential for women that have high aneuploidy rates.

Are there any PGT labs I should definitely avoid?

If you’re going to use Herasight’s ImputePGTA to get polygenic scores using regular old PGT-A data, there’s one particular PGT-A lab you should avoid using: Juno Genetics.

Juno not only has the lowest coverage depth of any PGT-A company (meaning imputation is less accurate), but they have a history of refusing data requests by customers and violating HIPAA rules around this. This will likely change at some point in the future (the law is actually pretty clear that Juno needs to provide raw data to customers, and they definitely are violating HIPAA with their current actions). But this may take some time to work its way through the legal system.

In the meantime, avoid using Juno at all costs.

How did you make the gain calculator?

I combined Herasight's IVF calculator with their gain calculator and pulled numbers on predictor quality from Genomic Prediction and Orchid from their most recently released paper (hat tip to Spencer Moore and the rest of the team for these numbers). I also used some unreleased numbers from Herasight.

In all cases, I've tried to get the most up-to-date numbers I could for all companies. In the case of Genomic Prediction, this process was very frustrating. The company claims to have better predictors than the ones shown in their published paper, but after several back-and-forth discussions with them, they decided it wasn't worth their time to even tell me the predictor performance, let alone validate it.

So I've defaulted to using the latest available numbers from their publications. Maybe they have better numbers internally, but they don't seem to consider it a high priority to communicate this to customers.

Will I get the same gain from embryo selection if I'm already smart, or already tall?

Basically yes. Counter-intuitively, your starting point for any continuous trait is almost irrelevant to the expected gain. Explaining exactly why that happens is perhaps beyond the scope of this last minute revision at 11:30 PM, but it's related to the number of genetic variants involved in the trait. Even very smart or very tall people only have a few hundred more "tall" or "smart" variants when compared to the average person. But there are over ten thousand such variants, so there's headroom in both directions.

The same can be said for almost any continuous trait.

Note, there is in fact reversion to the mean for these complex traits, so while you'll still see the same gain relative to an average child, you won't see the same gain relative to the parents; if the parents are super smart, even a screened embryo may be less smart than them in expectation (ditto for other continuous traits)

Mean reversion is a whole can of worms, but the mean you're reverting to is basically the mean of your grandparents, great-grandparents etc. If by some miracle everyone in your family tree going back a few generations was as great as you, then your children wouldn't see any mean reversion.

What else will I be able to select for in the future?

At some point in the next year or so it will likely be possible to nudge personality traits like extraversion and neuroticism, though the predictors for these traits are going to take another few years to catch up with predictors for disease and intelligence. You’ll also likely be able to select against severe depression risk later in 2026.

These sorts of traits are pretty interesting for singularity-pilled people because they are the sort of thing that could continue to have significance, even if artificial superintelligence is right around the corner.

We don’t yet have good predictors for athletic performance, or physical appearance beyond height and eye/hair color. More abstract but important traits like “courage”, “charisma” and “tendency to behave pro-socially” are also at least a few years off.