Astral Codex Ten

Aaron

21 min read

Aaron

Article

Aaron is a recurring person in the Astral Codex Ten archive, appearing 7 times across 7 issues between December 07, 2023 and February 05, 2026. The archive places it in contexts such as “Aaron heard this story and figured that brash Silicon Valley biotech might be able to find an alternative”; “two months ago, when Aaron finally synthesized it and tried it on himself”; “Aaron probably has already passed his period of highest BCS3-L1 activity”. It most often appears alongside FDA, Lumina, Sam Altman.

Metadata

  • Category: People
  • Mention count: 7
  • Issue count: 7
  • First seen: December 07, 2023
  • Last seen: February 05, 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.

Defying Cavity: Lantern Bioworks FAQ
December 07, 2023 · Original source
I talked to Lantern founder Aaron Silverbook to get an idea of how this works, both in a biological and an economic sense. Aaron was very knowledgeable and forthcoming, although he uses the phrase “YOLO” somewhat more often than most biotech founders. This post isn’t a verbatim interview transcript, just a writeup of what I learned based on his answers.
It lacks a peptide that its species usually uses to arrange gene transfers with other bacteria. The antibiotic helps it win the Darwinian competition in your mouth to become King Of The Oral Bacteria. The alcohol metabolism means it won’t produce lactic acid (and so won’t cause tooth decay). The peptide knockout prevents it from transferring genes back and forth with other bacteria that might either inactivate it or leak its advantage. 1.1: Where did this come from? Who invented it? Professor Jeffrey Hillman of the University of Florida. In 1985, he was surveying the microorganisms on his graduate students’ teeth (as you do). One grad student had an unusual strain of S. mutans with a natural version of mutations 1 and 2 (it produced mutacin-1140, and was resistant to it). Hillman realized the potential, and spent the next few decades adding mutations 3 and 4 and testing the results. 1.2: So how did it end up with a tiny startup in 2023? Professor Hillman started a company “Oragenics” and applied for FDA approval. The FDA demanded a study of 100 subjects, all of whom had to be “age 18-30, with removable dentures, living alone and far from school zones”. Hillman wasn’t sure there even were 100 young people with dentures, but the FDA wouldn’t budge from requiring this impossible trial. Hillman gave up and switched to other projects (including an intranasal COVID vaccine!) Aaron heard this story and figured that brash, move-fast-and-break-things Silicon Valley biotech might be able to find an alternative route to commercialization. The strain was off-patent, so he first tried to synthesize it himself from the clues in Hillman’s published papers. When that didn’t work, he made a deal with Oragenics for 10% of profits in exchange for samples and the full recipe. 2: How do you use it? To apply, you brush your teeth with a special pumice-based product that removes your existing tooth bacteria, then swab it on with a q-tip. One dose is sufficient; once you use it, it’s in your mouth approximately forever. 2.1: As users kiss their loved ones, who kiss others in turn, will this spread exponentially and take over the world? There was originally some concern about this, but no. Remember, the original bacterium was found in the wild, in a random grad student’s mouth forty years ago. There must be thousands of people walking around with various naturally-occurring BCS3-L1-like things. So probably this isn’t a risk for some kind of weird pandemic. Existing mouth bacteria have fortified their position and have a strong home field advantage. This is why you need to brush your teeth with the special product to apply Lumina. Lantern’s safety documents note that couples who kiss constantly do end up with similar oral microbiomes. So maybe enough kissing - especially kissing just after a dental cleaning when your existing bacteria are at their weakest - could spread the strain accidentally, very slowly. This rate of spread would be comparable to the rate of spread of every other mouth bacterium. 2.2: When a user kisses their newborn baby, will it spread to the baby? Okay, this one is true. Babies have no existing mouth bacteria, and get theirs from their parents’ kisses. Not necessarily their first kiss as a newborn (newborns have no teeth, and BCS3-L1 needs teeth to live), but their first kiss after teeth grow in. If you get this, you’re probably getting it for your whole future family line. 2.3: If you wanted to get rid of it, could you? Some kind of extreme course of oral antibiotics that nukes everything growing in your mouth would probably eradicate BCS3-L1, but this hasn’t been tested and would have side effects. 3: Is it dangerous to have bacteria secreting an antibiotic in your mouth? Does this mean you’re on a weak antibiotic all the time? There are already bacteria secreting antibiotics in your mouth. Microbes are in constant war with other microbes, and antibiotics are one of their favorite weapons - remember, penicillin comes from a fungus. Because bacteria secrete just enough antibiotic to clear their local area, these are tiny quantities, much less than you’d get from taking a medical-grade antibiotic pill. Lantern says the levels of mutacin-1140 dilute to irrelevance “tens of microns” away from the secreting bacteria. In any case, it’s a weak antibiotic that doesn’t survive the digestive tract (Hillman originally hoped to market the antibiotic too, but found it didn’t get absorbed and broke down too quickly). Neither the grad student with the original strain nor any of Hillman’s test subjects had any noticeable health issues. See also Lantern’s Safety Review FAQ. 3.1: Is it bad to disrupt your normal mouth microbiome? When talking about BCS3-L1 “taking over” the mouth, this just means it takes over the streptococcus mutans niche. There are still other bacteria and fungi in the mouth. The mutacin antibiotic might still disrupt these other bacteria (probably not fungi). But strains like BCS3-L1 already exist in the wild (eg the original grad student), and lots of bacteria and fungi secrete antibiotics, so it doesn’t seem like having mutacin-secreting organisms in your mouth makes you some extreme oral microbiome outlier. If you eat a normal Western diet, your mouth microbiome is already pretty far from the design specs, and it’s unclear if using Lumina makes things worse. 3.2: Will the other bacteria develop resistance to the antibiotic? Mutation 4 prevents BCS3-L1 from “leaking” its own resistance. Although in theory other bacteria could develop resistance, mutacin-1140 is a hard antibiotic to develop resistance to, and the other bacteria would have to do it in the short period before BCS3-L1 kills them off and establishes its own home field advantage. In practice, Professor Hillman found that BCS3-L1 remained dominant over many years and nothing developed resistance to it. Even if a mutacin-resistant strain does develop in one person’s mouth, it will have a hard time getting to anyone else’s mouth, so widespread immunity is unlikely. 4: Is it dangerous to have bacteria secreting alcohol in your mouth? Will you get drunk? Most people already have some alcohol-secreting bacteria in their bodies. (there’s a condition called auto-brewery syndrome where those bacteria get out of control and produce enough alcohol to make someone drunk. It’s vanishingly rare in real life, but more common in the legal system: “You gotta believe me, Officer, it was just auto-brewery syndrome!”) The average person has enough of these bacteria in their gut to have a natural blood alcohol level - even after zero drinks - of about 0.1 mg/dl. Under pessimistic assumptions, BCS3-L1 will add another 0.2 mg/dl, bringing the total to 0.3. This is still a pretty normal number that some people have naturally (it would bring the average customer from the ~50th to the ~80th percentile of natural blood alcohol). It’s also far from the usual threshold for feeling tipsy (30 mg/dl) or too drunk to drive (80 mg/dl). Under more realistic assumptions, the amount of alcohol produced by BCS3-L1 probably isn’t significant even by the very low standards of natural blood alcohol concentrations. 4.1: Are there some unusual scenarios where this amount of alcohol might matter? I don’t think Lantern has studied Breathalyzers. Since the alcohol is directly in your mouth, it might have disproportionate effect on a Breathalyzer compared to alcohol in your blood. I think it’s probably still too low to matter, but this is a wild guess. There is conjecture that “non-alcoholic steatohepatitis”, a liver disease in which non-alcoholics get the same kind of liver damage that alcoholics usually get, might be associated with endogenous blood alcohol in the high normal range. If I’m understanding this paper right, it’s probably because the gut produces levels of alcohol consistent with auto-brewery syndrome, the liver goes into overdrive and metabolizes it away (prevents auto-brewery syndrome from developing), but the liver is damaged in the process the same as if it had to go into overdrive to metabolize normal binge drinking. Since BCS3-L1 produces much less alcohol than auto-brewery, I think it wouldn’t cause non-alcoholic steatohepatitis, even though it might produce final blood alcohol levels similar to those associated with the condition. I was originally worried that Lumina might activate Antabuse, an anti-alcoholism drug that prevents drinking by causing a very unpleasant (sometimes dangerous) reaction to ethanol. There are some past cases of Antabuse being activated by really trivial quantities, like the alcohol in a chicken marsala dish or a mouth wash. But no, I think BCS3-L1 is less than this too. Chicken marsala can contain several grams of alcohol per serving, but BCS3-L1 probably only produces a few milligrams per day. If you swallow 1/10 of your mouthwash, that’s about 200 mg of alcohol - again, BCS3-L1 is probably only a few milligrams a day. Antabuse usually activates around a BAC of 5 mg/dl; BCS3-L1 only gives you a BAC of about 0.3 mg/dl. Again, this really is a tiny amount of alcohol. There might be other edge cases like these. Lantern offers a $100 bounty to anyone who can come up with one they haven’t thought of yet (and sometimes extra if you’re willing to help them research them). 4.2: Has anyone tested this in real life? As mentioned before, the mutacin-releasing strain (with mutations 1 and 2) exists in the wild and was extensively tested by Professor Hillman. The full strain with all four mutations has undergone some testing by Dr. Hillman, but nobody had officially infected themselves with it until two months ago, when Aaron finally synthesized it and tried it on himself. He says he’s usually “a lightweight” as far as alcohol goes, and hasn’t felt any different over the past two months. When first infected, BCS3-L1 makes up almost 100% of the microbiome (because you deliberately removed all your other bacteria, then infected yourself with it). Over time, other bacteria creep back in; over an even longer period (years?), BCS3-L1 reclaims lost territory and reaches a steady state. But the point is that Aaron probably has already passed his period of highest BCS3-L1 activity, and felt nothing. My wife infected herself about a month ago, and I haven’t noticed her having worse judgment or becoming more impulsive. But at baseline she was the sort of person who would infect herself with an untested genetically-modified bacteria strain, so there might be floor effects. 5: What’s the plan to sell Lumina? The plan is: Phase 1: (January 2024) Sell to biohackers in Prospera for $20,000.
Inline links: an intranasal COVID vaccine, made a deal, recipe, To apply, Safety Review FAQ, auto-brewery syndrome, in the legal system, this paper, can contain, https://substackcdn.com/image/fetch/$s_!My2N!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9200a48f-2549-40ba-9b4a-73c3e19fe93a_569x393.png
By good luck, Prospera will soon be hosting a two month super-conference of biotech and crypto entrepreneurs/enthusiasts. Aaron thinks they sound like the sorts of people who might want an experimental cavity-preventing bacterium and have $20,000 to spend, so he’s hoping to sell at least twenty doses. But also, anyone else who wants the product can go on a medical tourism trip to Prospera and get it in their experimental-treatment clinic.
Inline links: a two month super-conference, experimental-treatment clinic
Open Thread 306
December 11, 2023 · Original source
1: Comments of the week: Timothy Johnson tries to explain why Taylor Swift is such a big deal; Waldo explains why regulated businesses might sue their regulators even when trying to stay on their good side; and Scott Aaronson sets the record straight on his beliefs about AI risk. And I previously said I couldn’t find the source of a poll claiming that the median American estimated a 26% chance AI would kill all humans, but an alert reader found it here. Remember that ordinary people aren’t good at asserting probabilities, and also that medians don’t always present the full picture; see the link for more details.
Inline links: why Taylor Swift is such a big deal, why regulated businesses might sue their regulators even when trying to stay on their good side, sets the record straight on his beliefs about AI risk, found it here
4: Since I was pretty gung ho about the Lumina tooth probiotic, I want to link the good criticism I found as a counterbalance (without necessarily endorsing it). Here’s someone from Hacker News doubting that it will colonize the mouth (or do much if it does) - though see comments below. Here’s an endodontist talking about how hard it is to study this or get any evidence that it works. Some other people pointed out that the graph on the post shows only 50% colonization after one year; Aaron says he has other information showing it eventually reaches near 100% colonization and he’ll get that to me soon. Some people were extremely skeptical about whether any of this was even real, so here’s an NYT article about the original Hillman research from 2004 that will hopefully put those doubts to rest. I agree that there isn’t proof of efficacy and it will be hard to prove that, but I think the suggestive evidence compares well to other supplements I respect (albeit not $20K supplements), and no one in the comments had a good story for why it would cause harm - so I chose to take the free sample. I’ll let you know if anything terrible happens to me! Till then, you can also check the prediction markets.
Inline links: someone from Hacker News doubting, an endodontist talking about, here’s an NYT article about the original Hillman research from 2004, the prediction markets
Updates on Lumina Probiotic
April 16, 2024 · Original source
I got this information from (company CEO) Aaron, who says he got it from (original inventor) Jeffrey Hillman.
Or someone in the Dr. Hillman → Aaron → me chain mixed up details of the three trials into a mishmash with some characteristics of each.
The rats with the new strain (BCS3-L1) got only 1/3 the normal rats’ “caries score”. But they didn’t get a score of zero. So maybe claims like “BCS3 represents a complete cure for cavities” are overblown. Why didn’t rats with the new strain get zero dental caries? Bacteria other than S. mutans can also cause cavities, so maybe it’s one of those. Rat trials are famous for results that don’t replicate in human trials, so take these with a grain of salt. 3: What did the latest colonization studies show? Aaron was able to retest six people who got free samples in December. Four of those people still have the bacterium. The other two don’t. Of the two failures, one had an active cavity at the time the strain was applied (which interferes with the oral microbiome), and the other had his wisdom teeth removed (which involves rinsing the mouth with strong antiseptics). Aaron hopes this shows the strain will stick around in most normal situations (though the failure in the presence of active cavities is disappointing). 4: Any new concerns about side effects? In my original post, I mentioned the possibility that this would set off Breathalyzers. Lantern was able to test this, and proved that it wasn’t a problem. Yesterday, Lao Mein suggested on Less Wrong that it might raise oral cancer risk - their post focused on people with ALDH deficiency (most common in Asians) but the calculations are too vague to be sure exactly which groups should and shouldn’t worry. This is less than 24 hours old, the company hasn’t replied yet, and is still developing. I’ll try to update people if anyone gets more clarity on this. Someone on the post mentioned that they’ve gotten worse hangovers since using the product, maybe because the constant trickle of alcohol changed the way gut flora metabolize it. 5: Any other meaningful results since the samples? Cremieux says his breath smells better. Some people have objected to this claim on the grounds that it takes ~12 months before the bacterium has colonized your mouth. One of the figures in my earlier post suggested that the bacterium might start strong, retreat for a while, and then take 12 months to fully colonize, so that might potentially explain his findings. But also, is it biologically plausible that this prevents bad breath? My impression was that bad breath came from other bacterial byproducts besides lactic acid. It might be possible in theory that the same metabolic changes that switch lactic acid to alcohol disrupt these other byproducts, but it seems kind of unlikely. An alternate explanation is that, in order to apply this product at all, you need to do a dentist-style teeth cleaning that kills your previous mouth bacteria. Maybe that improves the bad breath regardless of whether you add the Lumina afterwards? Some other people have said their mouth feels fresher or something, but realistically all of this is overwhelmingly likely to be placebo. 6: Do I “endorse” Lumina? Richard Hanania has a post about how he trusts Lumina because I’ve endorsed them. It’s extremely kind and I appreciate his respect. But also, the most I said in the original post was that I was still debating whether or not to get the treatment. My real opinion, as precisely as I can express it, is: Advance of approximately the same magnitude as fluoride: 5%
Inline links: Bacteria other than, Lao Mein suggested on Less Wrong that it might raise oral cancer risk, says his breath smells better, has a post about how he trusts Lumina because I’ve endorsed them
Links for May 2024
May 29, 2024 · Original source
Aaron, CEO of Lumina, responded with an email asking him to take it down; it technically did not threaten him with a defamation suit, but had strong vibes to that effect. Then another Lumina employee independently sent an email actually threatening to sue for defamation. Trevor very reasonably published both emails on his blog, people very reasonably turned against Lumina, and I understand Lumina will have a statement or something soon.
Inline links: very reasonably published both emails on his blog
Meetups Everywhere 2024: Times & Places
August 29, 2024 · Original source
Contact: Aaron Contact Info: ironlordbyron[a t]gmail[d ot]com Time: Sunday, October 13th, 04:00 PM Location: Davanni's Pizza: 41 Cleveland Ave S, St Paul, MN 55105 Coordinates: https://plus.codes/86P8WRQ6+XX Group Link: Discord link for the MSP ACX meetup group: https://discord.gg/m2xJcuC937 Notes: I'll be providing pizza! Vegans are free to bring their own food (Davanni's selection here isn't great); I'll be getting a vegetarian and gluten-free pizza along with other kinds of pizza. RSVPs on lesswrong would be great.
Inline links: https://plus.codes/86P8WRQ6+XX, https://discord.gg/m2xJcuC937
Press Any Key For Bay Area House Party
July 21, 2025 · Original source
“So what are you going to do about it?” asks one of the other EAs. His nametag identifies him as an Aaron.
“Too late,” says Aaron. “I heard Zuck offered him fifty billion to work at Meta AI.”
You turn to Aaron. “What do you work on?”
Links For February 2026
February 05, 2026 · Original source
55: Related: there’s a common argument that maybe these statistics are wrong and biased, and crime rates have actually gone up. It goes: most crimes are plagued by reporting bias - if crime gets too bad, people simply don’t bother telling the police or other data-collecting bodies. The only crime that isn’t like this is murder - everyone notices a missing/dead human, and the police have to investigate all of them. So the only trustworthy statistic is the murder rate. Murders have gone down, but this is an artifact of improved trauma care saving many victims’ lives; if trauma care has gotten twice as good, then the apparent number of murders will halve even for the same amount of crime. If you adjust the apparent murder rate for the improvement in trauma care, real murder rates may have doubled or even more. Aaron Zinger investigates at the bottom of this post and disagrees; he says that murders averted by trauma care should show up as assaults, but that assaults have declined at almost the exact same rate as murders, suggesting a genuine decrease in people attacking one another, regardless of outcome - otherwise, it would be too much of a coincidence for the (trauma-care-induced) decline in the murder rate to exactly correlate with the (recording-bias-induced) decline in the assault rate. But how could improved trauma care not be biasing murder data collection? Aaron argues that would-be murderers have adjusted by trying harder to kill their victims (eg leaving them for dead vs. shooting them again to be sure). I’m a little skeptical of this (does the average murderer really calibrate their murder severity to the trauma care level? Aren’t many murders in very fast attacks where the murderer doesn’t get to choose how many shots/stabs to land?) and would welcome more research on this topic.
Inline links: Aaron Zinger investigates at the bottom of this post