Astral Codex Ten

nematodes

19 min read

nematodes

Article

nematodes is a recurring concept in the Astral Codex Ten archive, appearing 2 times across 2 issues between November 17, 2021 and February 12, 2026. The archive places it in contexts such as “nematodes, trematodes, all sorts of worms”; “anchor assumed away all animals other than nematodes”. It most often appears alongside 2010 kink, 2024 kink, ACE-2 receptor.

Metadata

  • Category: Concepts
  • Mention count: 2
  • Issue count: 2
  • First seen: November 17, 2021
  • Last seen: February 12, 2026

Appears In

Source Context

Recovered passages from the original issue text. When the raw archive preserved outbound links inside the source passage, they are listed directly under the quote.

Ivermectin: Much More Than You Wanted To Know
November 17, 2021 · Original source
Here’s the prevalence of roundworm infections by country (source). But alongside roundworms, there are threadworms, hookworms, blood flukes, liver flukes, nematodes, trematodes, all sorts of worms. Add them all up and somewhere between half and a quarter of people in the developing world have at least one parasitic worm in their body. Being full of worms may impact your ability to fight coronavirus. Gluchowska et al write: Helminth [ie worm] infections are among the most common infectious diseases. Bradbury et al. highlight the possible negative interactions between helminth infection and COVID-19 severity in helminth-endemic regions and note that alterations in the gut microbiome associated with helminth infection appear to have systemic immunomodulatory effects. It has also been proposed that helminth co-infection may increase the morbidity and mortality of COVID-19, because the immune system cannot efficiently respond to the virus; in addition, vaccines will be less effective for these patients, but treatment and prevention of helminth infections might reduce the negative effect of COVID-19. During millennia of parasite-host coevolution helminths evolved mechanisms suppressing the host immune responses, which may mitigate vaccine efficacy and increase severity of other infectious diseases. Treatment of worm infections might reduce the negative effect of COVID-19! And ivermectin is a deworming drug! You can see where this is going… The most relevant species of worm here is the roundworm Strongyloides stercoralis. Among the commonest treatments for COVID-19 is corticosteroids, a type of immunosuppresant drug. The types of immune responses it suppresses do more harm than good in coronavirus, so turning them off limits collateral damage and makes patients better on net. But these are also the types of immune responses that control Strongyloides. If you turn them off even very briefly, the worms multiply out of control, you get what’s called “Strongyloides hyperinfection”, and pretty often you die. According to the WHO: The current COVID-19 pandemic serves to highlight the risk of using systemic corticosteroids and, to a lesser extent, other immunosuppressive therapy, in populations with significant risk of underlying strongyloidiasis. Cases of strongyloidiasis hyperinfection in the setting of corticosteroid use as COVID-19 therapy have been described and draw attention to the necessity of addressing the risk of iatrogenic strongyloidiasis hyperinfection syndrome in infected individuals prior to corticosteroid administration. Although this has gained importance in the midst of a pandemic where corticosteroids are one of few therapies shown to improve mortality, its relevance is much broader given that corticosteroids and other immunosuppressive therapies have become increasingly common in treatment of chronic diseases (e.g. asthma or certain rheumatologic conditions). So you need to “address the risk” of strongyloides infection during COVID treatment in roundworm-endemic areas. And how might you address this, WHO? Treatment of chronic strongyloidiasis with ivermectin 200 µg/kg per day orally x 1-2 days is considered safe with potential contraindications including possible Loa loa infection (endemic in West and Central Africa), pregnancy, and weight <15kg. Given ivermectin’s safety profile, the United States has utilized presumptive treatment with ivermectin for strongyloidiasis in refugees resettling from endemic areas, and both Canada and the European Centre for Disease Prevention and Control have issued guidance on presumptive treatment to avoid hyperinfection in at risk populations. Screening and treatment, or where not available, addition of ivermectin to mass drug administration programs should be studied and considered. This is serious and common enough that, if you’re not going to screen for it, it might be worth “add[ing] ivermectin to mass drug administration programs” in affected areas! Dr. Avi Bitterman carries the hypothesis to the finish line: First two images are with all relevant studies; second two are a sensitivity analysis that removes some of the most dubious. The good ivermectin trials in areas with low Strongyloides prevalence, like Vallejos in Argentina, are mostly negative. The good ivermectin trials in areas with high Strongyloides prevalence, like Mahmud in Bangladesh, are mostly positive. Worms can’t explain the viral positivity outcomes (ie PCR), but Dr. Bitterman suggests that once you remove low quality trials and worm-related results, the rest looks like simple publication bias: This is still just a possibility. Maybe I’m over-focusing too hard on a couple positive results and this will all turn out to be nothing. Or who knows, maybe ivermectin does work against COVID a little - although it would have to be very little, fading to not at all in temperate worm-free countries. But this theory feels right to me. It feels right to me because it’s the most troll-ish possible solution. Everybody was wrong! The people who called it a miracle drug against COVID were wrong. The people who dismissed all the studies because they F@#king Love Science were wrong. Ivmmeta.com was wrong. Gideon Meyerowitz-Katz was…well, he was right, actually, I got the worm-related meta-analysis graphic above from his Twitter timeline. Still, an excellent troll. Also, the best part is that I ignorantly asked, in my description of Mahmud et al above: And it was! It was a fluke! A literal, physical, fluke! For my whole life, God has been placing terrible puns in my path to irritate me, and this would be the worst one ever! So it has to be true! The Scientific Takeaway About ten years ago, when the replication crisis started, we learned a certain set of tools for examining studies. Check for selection bias. Distrust “adjusting for confounders”. Check for p-hacking and forking paths. Make teams preregister their analyses. Do forest plots to find publication bias. Stop accepting p-values of 0.049. Wait for replications. Trust reviews and meta-analyses, instead of individual small studies. These were good tools. Having them was infinitely better than not having them. But even in 2014, I was writing about how many bad studies seemed to slip through the cracks even when we pushed this toolbox to its limits. We needed new tools. I think the methods that Meyerowitz-Katz, Sheldrake, Heathers, Brown, Lawrence and others brought to the limelight this year are some of the new tools we were waiting for. Part of this new toolset is to check for fraud. About 10 - 15% of the seemingly-good studies on ivermectin ended up extremely suspicious for fraud. Elgazzar, Carvallo, Niaee, Cadegiani, Samaha. There are ways to check for this even when you don’t have the raw data. Like: The Carlisle-Stouffer-Fisher method: Check some large group of comparisons, usually the Table 1 of an RCT where they compare the demographic characteristics of the control and experimental groups, for reasonable p-values. Real data will have p-values all over the map; one in every ten comparisons will have a p-value of 0.1 or less. Fakers seem bad at this and usually give everything a nice safe p-value like 0.8 or 0.9.
Inline links: source, Gluchowska et al, the WHO, carries, https://substackcdn.com/image/fetch/$s_!xExE!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F5da21781-249c-4e59-b616-9f23d83cc044_2048x1184.jpeg, https://substackcdn.com/image/fetch/$s_!4SMr!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fdcd6e4b2-37f7-4602-93d5-2581c3b27a60_700x432.png, https://substackcdn.com/image/fetch/$s_!-6n2!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F7fd6e8f4-093e-4e02-bce7-363615146c9c_2228x1346.jpeg, https://substackcdn.com/image/fetch/$s_!CPZs!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2Fb0425847-198a-4bd3-a63b-149f15d147ba_700x432.png, https://substackcdn.com/image/fetch/$s_!H3rK!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F9972491b-25b0-4c06-8aca-86fce102ae63_666x147.png, even in 2014, The Carlisle-Stouffer-Fisher method
What Happened With Bio Anchors?
February 12, 2026 · Original source
Epoch/Croxton are current best estimates, and can probably fairly be read as the “real” answer against which Cotra and Davidson’s earlier guesses should be judged. All numbers are yearly multiples, so 1.4 means that willingness to spend grows 1.4x per year, ie 40%. Willingness To Spend: How much money are companies willing to spend on AI, in the form of chips and data centers? $/FLOP: How quickly do Moore’s Law, economies of scale, and other factors bring down the price of AI compute? Training Run Length: How long are companies spending on AI training runs for frontier models (instead of using those chips for smaller models, experiments, or consumer services)? Real Compute: The product of the three parameters above. Algorithmic Progress: How effectively do researchers discover new algorithms that makes training AIs cheaper and more efficient? Total Effective Compute: The product of real compute and algorithmic progress. So for example, the Epoch column’s 10.7x means that in any given year, you can train an AI 10.7x better than the last year, because you have 3.6x more compute available, and that compute is 3.0x more efficient. Cotra and Davidson were pretty close on willingness to spend and on FLOPs/$. This is an impressive achievement; they more or less predicted the giant data center buildout of the past few years. They ignored training run length, which probably seemed like a reasonable simplification at the time. But they got killed on algorithmic progress, which was 200% per year instead of 30%. How did they get this one so wrong? Here’s Cotra’s section on algorithmic progress: Algorithmic progress forecasts Note: I have done very little research into algorithmic progress trends. Of the four main components of my model (2020 compute requirements, algorithmic progress, compute price trends, and spending on computation) I have spent the least time thinking about algorithmic progress. I consider two types of algorithmic progress: relatively incremental and steady progress from iteratively improving architectures and learning algorithms, and the chance of “breakthrough” progress which brings the technical difficulty of training a transformative model down from “astronomically large” / “impossible” to “broadly feasible.” For incremental progress, the main source I used was Hernandez and Brown 2020, ”Measuring the Algorithmic Efficiency of Neural Networks”. The authors reimplemented open source state-of-the-art (SOTA) ImageNet models between 2012 and 2019 (six models in total). They trained each model up to the point that it achieved the same performance as AlexNet achieved in 2012, and recorded the total FLOP that required. They found that the SOTA model in 2019, EfficientNet B0, required ~44 times fewer training FLOP to achieve AlexNet performance than AlexNet did; the six data points fit a power law curve with the amount of computation required to match AlexNet halving every ~16 months over the seven years in the dataset.² They also show that linear programming displayed a similar trend over a longer period of time: when hardware is held fixed, the time in seconds taken to solve a standard basket of mixed integer programs by SOTA commercial software packages halved every ~13 months over the 21 years from 1996 to 2017.³ Grace 2013 (”Algorithmic Progress in Six Domains”) is the only other paper attempting to systematically quantify algorithmic progress that I am currently aware of, although I have not done a systematic literature review and may be missing others. I have chosen not to examine it in detail because a) it was written largely before the deep learning boom and mostly does not focus on ML tasks, and b) it is less straightforward to translate Grace’s results into the format that I am most interested in (”How has the amount of computation required to solve a fixed task decreased over time?”). Paul is familiar with the results, and he believes that algorithmic progress across the six domains studied in Grace 2013⁴ is consistent with a similar but slightly slower rate of progress, ranging from 13 to 36 months to halve the computation required to reach a fixed level of performance. Additionally, it seems plausible to me that both sets of results would overestimate the pace of algorithmic progress on a transformative task, because they are both focusing on relatively narrow problems with simple, well-defined benchmarks that large groups of researchers could directly optimize.⁵ Because no one has trained a transformative model yet, to the extent that the computation required to train one is falling over time, it would have to happen via proxies rather than researchers directly optimizing that metric (e.g. perhaps architectural innovations that improve training efficiency for image classifiers or language models would translate to a transformative model). Additionally, it may be that halving the amount of computation required to train a transformative model would require making progress on multiple partially-independent sub-problems (e.g. vision and language and motor control). I have attempted to take the Hernandez and Brown 2020 halving times (and Paul’s summary of the Grace 2013 halving times) as anchoring points and shade them upward to account for the considerations raised above. There is massive room for judgment in whether and how much to shade upward; I expect many readers will want to change my assumptions here, and some will believe it is more reasonable to shade downward. Cotra’s estimate comes primarily from one paper, Hernandez & Brown, which looks at algorithmic progress on a task called AlexNet. But later research demonstrated that the apparent speed of algorithmic progress varies by an order of magnitude based on whether you’re looking at an easy task (low-hanging fruit already picked) or a hard task (still lots of room to improve). AlexNet was an easy task, but pushing the frontier of AI is a hard task, so algorithmic progress in frontier AI has been faster than the AlexNet paper estimated. In Cotra’s defense, she admitted that this was the area where she was least certain, and that she had rounded the progress rate down based on various considerations when other people might round it up based on various other considerations. But the sheer extent of the error here, compounded with a few smaller errors that unfortunately all shared the same direction, was enough to throw off the estimate entirely. Since Cotra and Davidson were expecting AI to get 3.6x more effective compute each year, but it actually got 10.7x more, it’s no mystery why their timelines were off. When John recalculates Davidson’s model with Epoch’s numbers, he finds that it estimates AGI in 2030, which matches the current vibes. IV. With this information in place, it’s worth looking at some prominent contemporaneous critiques of Bio Anchors. Various people criticized Bio Anchors’ many strange anchors for how much compute it would take to produce AGI. For example, one anchor estimated that it would take 10^45 FLOPs, because that was how many calculations happened in all the brains of all animals throughout the evolutionary history (which eventually produced the human brain that AIs are trying to imitate). To make things even weirder, this anchor assumed away all animals other than nematodes as a rounding error (fact check: true!) All of these seemed to detract from the main show, an attempt to estimate the compute involved in the human brain. But even this more sober anchor was complicated by time horizons - it’s not enough to imitate the human brain for one second; AIs need to be able to imitate the human brain’s capacity for long-term planning. Cotra calculated how much compute AGI would require if it needed a planning horizon of seconds, weeks, or years. Thanks to METR, we now know that existing AIs have already passed a point where they can do most tasks that take humans seconds, are moving through the hour range, and are just about to touch one day. So the “seconds” anchor is ruled out. But it also seems unlikely that AGI will require years, because most human projects don’t take years, or at least can be split into tasks that take less than one year each (intuition pump: are we sure the average employee stays at an AI lab for more than a year? If not, that proves that a chain of people with sub-one-year time horizons can do valuable work). The AI Futures team guessed that the time horizon necessary for AIs to really start serious recursive self-improvement was between a few weeks and a few months (though this might look like a totally different number on the METR graph, which doesn’t translate perfectly into real life). If this is true, then all three anchors (seconds, hours, years) were off by at least an order of magnitude. But it turns out that none of this matters very much. The highest and lowest anchors cancel out, so that the most plausible anchor - human brain with time horizon of hours to days - is around the average. If you remove all the other anchors and just keep that one, the model’s estimates barely change. But also, we’re talking about crossing twelve orders of magnitude here. The difference between the different time horizon anchors doesn’t register much on that level, compared to things like algorithmic progress which have exponential effects. Maybe this is the model basically working as intended. You try lots of different anchors, put more weight on the more plausible ones, take a weighted average of each of them, and hopefully get something close to the real value. Bio Anchors did. Or maybe it was just good luck. Still hard to tell. Eliezer Yudkowsky argued that the whole methodology was fundamentally flawed. Partly because of the argument above - he didn’t trust the anchors - but also partly because he expected the calculations to be obviated by some sort of paradigm shift that couldn’t be shoehorned into “algorithmic progress” (like how you couldn’t build an airplane in 1900 but you could in 1920). As of 2026 - still before AGI has been invented and we get a good historical perspective - no such shift has occurred. The scaling laws have mostly held; whatever artificial space you try to measure models in, the measurement has mostly worked in a predictable way. There have really only been two kinks in the history of AI so far. First, a kink in training run size around 2010: Second, a kink in time horizons around 2024 and the invention of test-time compute: The 2010 kink was before Cotra’s forecast and priced in. The 2024 kink is interesting and relevant - but since it was on a parameter Cotra wasn’t measuring, and probably too small to show up on the orders-of-magnitude scale we’re talking about, it’s probably not a major cause of the model’s inaccuracy. Other things have been even more predictable: So Cotra’s bet on progress being smooth and measurable has mostly paid off so far. But Yudkowsky further explained that his timelines were shorter than Bio Anchors because people would be working hard to discover new paradigms, and if the current paradigm would only pay off in the 2050s, then probably they would discover one before then. You could think of this as a disjunction: timelines will be shorter than Cotra thinks, either because deep learning pays off quickly, or because a new paradigm gets invented in the interim. It turned out to be the first one. So although Yudkowsky’s new paradigm has yet to materialize, his disjunctive reasoning in favor of shorter-than-2050 timelines was basically on the mark. Nostalgebraist argued that Cotra’s whole model was a wrapper for an assumption that Moore’s Law will continue indefinitely. If it does, obviously you get enough compute for AI at some point, even if it requires some absurd process like simulating all 500 million years of multicellular evolution. I never entirely understood this objection, because - although Bio Anchors does depend on a story where Moore’s Law doesn’t break before we get the relevant amount of compute - this is only one of many background assumptions (like that a meteor doesn’t hit Earth before we get the relevant amount of compute). Given those assumptions, it does a useful not-just-assumption-repeating job of calculating when transformative AI will happen. As Cotra implicitly predicted, we seem on track to get AGI before Moore’s Law breaks down, and so Moore’s Law didn’t end up mattering very much. And if all of Cotra’s non-Moore’s-Law parameter estimates had been correct, her model would have given about the same timelines we have now, and surprised everyone with a revolutionary claim about the AI future. But Nostalgebraist added, almost as an aside: Cotra has a whole other forecast I didn’t mention for “algorithmic progress,” and the last number is what you get from just algorithmic progress and no Moore’s Law. So depending on how much you trust that forecast, you might want to take all these numbers with an even bigger grain of salt than you’d expected from everything else we’ve seen. How much should you trust Cotra’s algorithmic progress forecast? She writes: “I have done very little research into algorithmic progress trends. Of the four main components of my model (2020 compute requirements, algorithmic progress, compute price trends, and spending on computation) I have spent the least time thinking about algorithmic progress.” ...and bases the forecast on one paper about ImageNet classifiers. I want to be clear that when I quote these parts about Cotra not spending much time on something, I’m not trying to make fun of her. It’s good to be transparent about this kind of thing! I wish more people would do that. My complaint is not that she tells us what she spent time on, it’s that she spent time on the wrong things. Like Cotra herself, I think Nostalgebraist was spiritually correct even if his bottom line (about Moore’s Law) was wrong. His meta-level point was that a seemingly complicated model could actually hinge on one or two parameters, and that many of Cotra’s parameter values were vague hand-wavey best guess estimates. He gave algorithmic progress as a secondary example of this to shore up his Moore’s Law case, but in fact it turned out to be where all the action was. V. Those were the rare good critiques. The bad critiques were the same ones everyone in this space gets: You’re just trying to build hype.
Inline links: ”Measuring the Algorithmic Efficiency of Neural Networks”, ”Algorithmic Progress in Six Domains”, later research, fact check: true!, argued that, https://epoch.ai/blog/the-longest-training-run, https://www.lesswrong.com/posts/GAJbegsvnd85hX3eS/thoughts-on-extrapolating-time-horizons, https://epoch.ai/benchmarks/eci, argued that