Kevin Blake PhD

Every biologist knows the story of Fleming's chance discovery of penicillin. But is it true? Here, I write about inconsistencies in the canonical story, and explore a few alternative theories about what really happened in that St. Mary's lab in the summer of 1928.

@salonium Couldn't agree more! Science is done by very smart, meticulous ppl who know what they're doing. I think what often gets called 'accidents' are really just the 'unexpected'. But this shouldn't be unusual bc revealing the unexpected is EXACTLY what science does best.

I’m a disbeliever in accidental discoveries (at least, in biology). Whenever I’ve looked into one, the story turns out to be false. The most famous is penicillin – supposedly, the fungi wafted in through a window, fell into a petri dish of cultured staphylococci, and suppressed the bacteria’s growth. But in a recent article (asimov.press/p/penicillin-m…), @kevinsblake explains that doesn’t really work (grown staphylococci aren’t affected by penicillin; it only works if introduced before the bacteria begin growing); plus, Fleming’s notes on the discovery provide very little detail and the specific results he described couldn’t be replicated by other scientists (even though penicillin does work against staphylococci when introduced correctly.) There are more: Pasteur’s supposedly accidental discovery of a chicken cholera vaccine was more likely the result of systematic work by his then-assistant, Émile Roux. (jstor.org/stable/2332836…) And, as @NikoMcCarty writes, the discovery of GFP, nanopore sequencing, and optogenetics are also often described as accidents, but none of them happened that way either. nikomc.com/2026/04/01/opt… People love serendipity, so why am I bursting their bubble? I don’t think this is limited to accidental discoveries; I think many historical science anecdotes are highly embellished: - Edward Jenner didn’t deliberately expose a young boy with full-blown smallpox to test his vaccine (he used variolation); and he wasn’t the first to try using cowpox bsky.app/profile/scient… - Cobra catching bounties in British India didn’t lead to a rise in the number of snakebites, and there was only hearsay evidence that cobras were bred in response at all twitter-thread.com/t/169650089580… - Barry Marshall didn’t develop stomach ulcers from drinking a concoction of H. pylori (he did develop gastritis though…) cdn.centerforinquiry.org/wp-content/upl… - No one knows who actually found the highly-productive strain of penicillin on a cantaloupe, but it probably wasn’t 'Moldy Mary' scientificdiscoveries.ars.usda.gov/tellus/stories… But in this case it irks me for an additional reason – it gives the impression that innovation happens sporadically, by chance, when there are actually ways that we can systematically speed it up – such as better funding, institutions and incentives. So: are there any true accidental discoveries that hold up to scrutiny?

@LocasaleLab Punching down on trainees, who make <$35k/yr producing 2x the research for 1/4 the academic jobs, is about as low as you can go.

PhD students now can barely show up, do little to no work, have no interest in their thesis research, take on side jobs or excessive extracurriculars, and still draw a full salary and benefits. They can leave as they choose with a diploma, or be “fired” by being granted a degree. Many stay because they’re effectively being paid to remain living a college student life minus the responsibility of grades and exams — they even go to frat parties, date undergrads and live in dorms as they approach their 30s.

@NikoMcCarty MMs and other chemostats are wonderful little devices. I wonder if their disuse is an accident of history; they got overlooked during the big data, bulk sequencing craze of the ‘10s. As that seems to be petering out in favor of single-cell work maybe MMs will make a comeback?

There is a tiny device (first invented almost two decades ago) that lets you watch a single cell divide hundreds of times, under tightly-controlled conditions, and yet I almost never meet people who actually use it. This is a shame because there’s many interesting ideas that I think could be uniquely tested in such a device. Most experiments in biology are instead done in “bulk,” which is not a good way to deeply understand an organism. RNA-seq experiments, for example, are often done by growing lots of cells in a flask, exposing them to some chemical, and then killing all the cells, extracting their RNAs, and sequencing all of them together. The end result of this is just an average, and it completely obscures the messier (and more truthful) stochastic nature of life. But a Mother Machine, and other microfluidics tools like it, helps to solve this problem. It’s just a tiny device with a long trench through which nutrients flow. Cells travel down this trench and fall into little wells, etched perpendicularly to the main trench. Each of these wells is barely wide enough for a bacterium to fall inside. When a cell falls into a well, it keeps dividing and also has access to fresh nutrients, which are constantly pumped through. Waste molecules are continuously flushed out. As one cell divides into two, then four, then eight, and so on, some cells eventually extend out of the well entirely and get swept away with the current. The cell at the bottom, though, stays put and will keep dividing. Like many other “great” inventions, the Mother Machine was designed to answer a specific question. In a 2005 paper, some scientists claimed that, when a cell divides, whichever offspring inherits the “old pole” from the mother divides about 2 percent slower with each passing generation. (Said another way: When an E. coli cell divides, it builds a wall down the middle and cuts itself in two at that point. Each “daughter” cell has two ends; one end is made from the wall, and the other end is “old.” This old end gets passed down through the generations, again and again. The authors of the 2005 paper claimed that this constitutes a form of cellular aging, and that cells which inherit the old end are basically less fit than the other daughter cell. Provocative claim!) The Mother Machine was invented by a small team at Harvard to disprove this hypothesis. In the 2010 paper describing the device, they basically just strapped a camera to the microfluidics chip and recorded the growth rate for tens of thousands of cells, under constant nutrient conditions, for “hundreds of generations.” After tallying all the data, they concluded that “E. coli, unlike all other aging model systems studied to date, has a robust mechanism of growth that is decoupled from cell death.” In other words, growth does not slow down with age, and the 2005 claims were wrong. Other groups have since made modifications to the original device, using it to revisit classic experiments in molecular biology. In 2018, for example, a Swiss team modified the microfluidic chip to have two input channels, rather than just the one. With two ports, they could expose cells to different growth media at the same time. Or they could switch back-and-forth between the two conditions, or even expose cells to gradients of those conditions. Now, it has been known since the 1960s that E. coli cells prefer to eat glucose over lactose. When glucose runs out and only lactose is around, the cells activate their lac operon and begin making enzymes to digest it. Jacques Monod and François Jacob shared a Nobel Prize, in 1965, for figuring this out. But nobody had ever actually watched this “switch” at the level of single cells, under tightly-controlled conditions. But then the Swiss team made their modified Mother Machine. They flooded E. coli cells into the device, trapped them in wells, and switched the inputs between glucose and lactose every four hours. And what they found is that, when lactose comes in to replace glucose, every cell stops growing within three minutes. This outcome is extremely uniform! But the reverse — or the time it takes each cell to switch on its lac operon — is extremely variable. About one-fourth of cells start growing within 25-45 minutes, two-thirds start growing in one-to-three hours, and five percent of cells never grow again at all. By accounting for cells individually, in other words, the Mother Machine enabled these researchers to make observations which could never be made at the population-scale. And yet, Mother Machines still seem relatively rare! The blueprints are freely available online, but making these devices still requires an understanding of photolithography. The wells are only a micron wide, so they can’t be 3D-printed; one has to make a master mold using photomasks, cast PDMS in that mold, and then cure the polymer into that shape. The original specs only work for E. coli, too. If you wanted to study Bacillus or Caulobacter or yeast, you’d have to redesign the channels with different dimensions. A few companies sell Mother Machines, but they seem to be quite small. If Mother Machines did become widespread, though (maybe even cheap enough to ship in, say, a $100 kit for students) they could be used to run all kinds of interesting experiments. One idea is to combine a Mother Machine with a hypermutation tool, such that we can watch cells evolve in real-time. In a recent study, British scientists reported a way to do “highly mutagenic continuous evolution” in E. coli. The beauty of their tool is that it only requires two components: an error-prone DNA polymerase, and a replicon carrying a gene of interest. The error-prone polymerase, which introduces about one mutation per 1,000 bases every ten generations, only copies the DNA on the replicon; it doesn’t touch the host genome. One could take a gene encoding antibiotic resistance (against molecule X) and clone it onto the replicon, transform the whole thing into E. coli, and trap the cells inside a Mother Machine. Then, by exposing the cells to increasing levels of antibiotic Y, one could watch in real time as cells mutate their resistance gene and, perhaps, hit upon a solution that confers resistance against both molecules. This would be a way to study how cells evolve resistance autonomously, at the single-cell level. Another idea is to use Mother Machines to study how perturbations change a cells’ transcriptome in real-time. Felix Horns (previously in Michael Elowitz’s group at Caltech, now at Arc Institute) created an RNA Exporter tool. The gist is that genes encoding virus-like particles are placed into cells and, when these particles get made, they latch onto RNA molecules and physically carry them out of the cell. Cells are effectively engineered to export their own RNA. My understanding is that RNA Exporters are relatively unbiased, meaning they have a roughly equal chance of grabbing onto any RNA molecule. The molecules that get carried from the cell, then, are representative of the transcriptome as a whole. If cells carrying RNA Exporters were studied in a Mother Machine, it might be possible to perturb them and measure their transcriptional responses in real time — rather than the classical approach of perturbing millions of cells at once in a flask and doing RNA-seq on the entire population to collect average results. A third idea is to collect single-cell observations to train a predictive model for molecular burden. Any time we engineer an organism to carry new genes, we are forcing it to execute a function it wouldn’t normally do, thus draining resources that would otherwise go toward growth, DNA repair, and so on. Perhaps we could take 100+ plasmids, each carrying a fluorescent protein, and clone all of them into the same strain of E. coli. Then we could study each strain inside a Mother Machine, carefully quantifying growth rates and fluorescence levels, to map out the full distribution of outcomes for a given plasmid. If we did this enough times (hopefully with some kind of automated data pipeline) we could collect a huge dataset. The resulting model could also help bioengineers design constructs that impose less of a burden on living cells. I’m not entirely sure why we’re not seeing more of these ideas implemented, or why bioengineers still haven’t fully embraced single-cell experiments. Every university should have a microfluidics facility making custom devices, but I’ve only visited a few of them. Most experiments are still done in bulk, using orders-of-magnitude more cells and reagents than microfluidics would require; and usually the results are less representative of ground truth, too! It’s a shame, because one of the beautiful things about biology is that each cell is unique and lots of molecular phenomena are highly stochastic, following a distribution of outcomes. Biology is fun because it is not deterministic; and that makes it both richer as a field of study but also more complicated as an engineering medium. A Mother Machine, and other tools like it, help us to actually see these distributions.

@Insane_Econ “Modern” is doing A LOT of heavy lifting

"Economics actually predates modern biology, and, indeed, modern biology was in a sense founded by the world's first professor of economics, Malthus." -Gordon Tullock, "Economics and Sociobiology: A Comment" (1977)

@FEhrsam @owl_posting Textbook explanations are oversimplifications precisely BECAUSE they’re made for children. Practicing scientists understand the complexity. They choose to use simplistic terms, even if they don’t capture the complexity, because it’s easier / faster. Especially for students.

The human body is a multivariate system that is complex beyond what the human mind can comprehend. Computers will be able to understand and aid the body in ways humans never could. Biology textbook explanations will be looked back on as child-like oversimplifications.

Asmiov Press has been THE platform to read and write about biology. Sad to see it go, even temporarily. But excited to see what @NikoMcCarty, @AlexandraBalwit and the team cook up next.

@salonium Evidence from evolution experiments with microbes, too: worksinprogress.co/issue/the-evol…

Excited to listen to this episode, but I disagree that Darwin's theory can't be decisively tested (as much as Newton's). I'd highly recommend the book 'Why Evolution Is True'. It describes how Darwin, and others, made a lot of predictions that were verified, with many different lines of evidence – including fossils, vestigial structures, biogeography, molecular evidence, natural selection in action, and speciation. Darwin was a great interlocutor but many of these predictions were tested in his work, or during his lifetime. And some of his auxiliary hypotheses did not hold up, and helped refine the theory. Anyway, it's a great book!

@arjunrajlab There are scores of scientific illustrators who would *love* to collaborate on such projects

The 20th century saw the style of scientific figures and writing go from whimsical to minimalist, the 21st century from minimalist to quasi-brutalist. Given that AI can mimic this new affect ~perfectly, perhaps we will see a return to whimsy to signify our humanity?

@PracheeAC Reading books by biologists from the 80s and earlier has been such an eye-opener. So much thought about scientific theory and history, which hardly gets a peep today.

Sometimes I wonder how so many PhDs have a massive blind spot for epistemology — because it’s not like we didn’t teach it. So I suspect it’s because we let careerism supplant intellectual curiosity and they no longer give a fuck

No maps gets to be called “smart” until it knows whether you’re on Lower vs. Upper Wacker

A non-hyped explainer of the “cell simulation” paper. The recent study about the “4D” simulation of a minimal cell has been getting a lot of attention on social media. Unfortunately, most posts about it have serious errors. I’ve seen people claim that the model simulates every chemical reaction in the cell, for example, which is not true. Some biomolecules and reactions *are* tracked individually in the simulation, including proteins and RNA (and ribosomes), and the chromosome. But the simulation does not track individual metabolites (like ATP or glucose), water, nucleotide precursors, lipds, and so on. These "other" molecules are represented, instead, as concentrations (using ordinary differential equations). But anyway, here goes my quick explanation: Researchers built a computational model that simulates roughly 100 minutes of biological time, or one cell division, for a single bacterial cell. Each simulation takes 4–6 days to run on two NVIDIA A100 GPUs, and the authors ran it 50 times in replicate. The cell simulation includes some elements of randomness, so each replication attempt leads to a slightly different outcome. When they plotted out these replicates and averaged results, they found that the model could predict a few things without being fitted to experimental data: The simulated cells “divided” every 105 minutes, on average, which matches experimental results; and the mRNA molecules had an average half-life of 3.63 minutes, which is roughly what we’d expect from experiments, too. The cell they are modeling is called JCVI-syn3A, and it is not a naturally-occurring organism. It’s a bacterium that has been engineered, over many years, to have a small genome. It only has 493 genes (compared to 4,000+ for E. coli), all of which are housed on a single chromosome. The Syn3A cell was made by taking a natural organism, called Mycoplasma mycoides, and then slashing out non-essential genes. Its entire proteome, transcriptome, and metabolism have been studied in depth, which is why it’s being used to build these whole-cell simulations. The actual *simulation*, though, is not a single thing! Instead, the authors wrote down all the “stuff” that happens inside a cell (transcription! translation! metabolism! lipid biosynthesis!) and decided which type of mathematical model would be best-suited to describe each thing. Some cell processes were modelled deterministically, others had “spatial” elements, and other parts were relatively random. More specifically, they used four different types of models to build this simulation: 1. A Reaction-Diffusion Master Equation, which was used to model the individual proteins, RNAs, and ribosomes. 2. A Chemical Master Equation, which was used to model things where spatial location doesn’t matter as much (it basically treats the whole cell as one mixed entity); including tRNA charging. 3. Ordinary Differential Equations, which you may be familiar with from Calculus class, were used to model changes in ATP concentration, lipids, and so on. 4. Brownian Dynamics, which simulated the chromosome as a physical chain of beads, where each bead represents 10 base pairs of DNA. The Reaction-Diffusion Equation works like this: Basically, they chopped up the entire digital cell into a 3D grid of cubes. Each cube measures 10 nanometers on each side. The whole cell is about 500 nanometers across, so there are tens of thousands of cubes in the cell's interior. (This is a useful way to coarse grain the simulation; if the cubes were smaller, the simulation would take much longer to run.) Each cube is a little box that contains some number of molecules. At every “step” in the simulation, only one of two things can happen to the molecules in each box: Either they react with a molecule in the same box, or they diffuse (“hop”) to an adjacent box. That’s it; the model is just rolling a die for each molecule at each time step in each box, and using those results to decide how each molecule changes over time. (The reason this spatial model is important is because biology only works if molecules physically bump into each other. And so this spatial grid means that, unlike simpler models, a protein actually has to “diffuse” across boxes in the cell to encounter its reaction partner; only then can it react and do something useful.) So anyway, each of these models is used to represent a different type of molecule. It’s not like there is a single, all-powerful simulation that they are running here; instead, they’re running these four models together, using a script that synchronizes their results with each other. The Reaction-Diffusion equation is the main part of the simulation. It takes time steps of 50 microseconds of biological time. Every 12.5 milliseconds of biological time — meaning every 250 RDME steps — the simulation pauses so that the other models can synchronize based on the latest state of the simulation. The Brownian Dynamics part runs on a completely separate GPU, and only updates every four seconds of biological time. So that's the gist here. But let's also be honest about what this simulation does NOT do: - It does not include polysomes, which are a cluster of ribosomes that all latch onto a single mRNA and translate at the same time. Polysomes are really common inside of cells, but this simulation assumes that each mRNA can only be translated by one ribosome at a time. - It does not include polycistronic transcription. In bacteria, genes are often grouped next to each other on the chromosome and thus “transcribed” (or turned into mRNA) all at once, together. The majority of genes in E. coli, for example, are arranged in these operons, and the authors of this paper acknowledge that many Syn3A genes are likely co-transcribed the same way. But the simulation doesn't capture it. - The authors manually tuned many parameters to get the model to make predictions that more closely resemble experiments. Earlier simulations were waaaayyyyy off from experimental results. For example, they adjusted the ratio of mRNA binding rates to ribosomes versus degradosomes because, in earlier simulations, mRNA was being degraded too quickly, before ribosomes could translate it, causing most proteins to be severely underproduced. - In the Brownian Dynamics model, the authors added a “fake” 12 pN physical force to push the two daughter chromosomes apart during division, because the real biological mechanism for chromosome partitioning in Syn3A is not known. - And some other things. That being said: This model is really cool! I love papers like this! I'm enamored by scientists who choose really difficult problems (like simulating an entire cell) and actually go after it and make progress! This paper is amazing because it shows us what we are able to simulate well, and what we don't yet understand, and to figure out which experiments we ought to perform to reconcile the two. So instead of framing this paper as "OH MY GOSH SCIENTISTS FIGURED OUT HOW TO SIMULATE AN ENTIRE CELL!" we should frame it as proof that there is still plenty of room at the bottom, many measurements to be made, and many avenues to explore as we seek to understand biology better.

@NikoMcCarty I've often dropped essays between 2 and 3 because the ideas loses its glitter when fleshed out to >500 words. BUT then have an adjacent idea weeks later, and the two together make a complete story.

It is the final essay that counts in writing. Readers only see the final version. Hundreds of other drafts and ideas remain squirreled away. So if we assume that writing an essay requires six steps: 1. Having an idea. 2. Writing down the idea. 3. Finishing a first draft. 4. Getting feedback on that draft. 5. Editing. 6. Publishing. ...then even if you had a 90% chance of completing each of these steps individually (which is remarkably high), the odds of going from idea to publication is only about 50%. This is because failure compounds (90%, multiplied six times, is 53%), and you must complete *every step* to get to the final essay. In other words, the writing process can break down at many different stages. Many ideas that should be written never are. Gwern calls this “the pipeline” and uses it to explain "the log-normal distribution of outputs” that we see from writers, where some are highly prolific and others rarely publish at all. If you want to write more, it’s worth thinking about which step is hardest for you. I’ve talked to lots of writers who create drafts but never publish because they think strangers on the internet will call them an idiot. Other writers have a million ideas, many of which *should* be essays, but never even jot them down. My own bottleneck is #3. I have dozens of unfinished drafts right now, but I jump around too much and get distracted by new ideas. If even one step has a low probability of success, it tanks the odds that you'll finish an essay. For example, if I only had a 50% chance of finishing step #3 (instead of 90%), then my odds of finishing the essay would drop from 50% to 30%. TL;DR: Find your writing bottleneck! And reach out to me if you'd like any advice or help: nsmccarty3@gmail.com

@AlexandraBalwit @AsimovPress Congrats! Looking forward to seeing the writing you do there.

After 2+ incredibly rewarding years at @AsimovPress, I am moving on. This coming week, I will be joining the editorial team at Anthropic. Finally, my penchant for em-dashes will meet a welcome embrace. I couldn't be more grateful or more excited for what's next.

NEW: Agar is a ubiquitous part of research labs, used to grow microbes to make vaccines & antibiotics. The person who first discovered its "superior features" for microbiology was Fanny Hesse, a “German housewife” whose contributions to science have mostly been forgotten.

@ATinyGreenCell Most days, feels like the only science on here are AI bros saying bio is dead. It’s exhausting.

I keep seeing msgs from academic friends saying "im not on X anymore; too much hate" and its just so saddening to see one of the best social networks for science and its communication collapse like this. BlueSky algo is so weak by compari I miss Old Science Twitter so much :((((

@NikoMcCarty @AsteraInstitute Sounds like a dream job. Congrats!

Personal update: I moved to the Bay Area to join @AsteraInstitute as an independent fellow. I'll spend my time writing weekly essays, researching, hosting dinners, starting a podcast, finishing my book, etc. Please invite me to things!

@kmcannon So do it. Produce two doctoral dissertations a day, with zero effort. Single-handedly quadruple human knowledge in a year. If this really is true, what's stopping you? What's stopping everyone else?

There are PhDs being handed out each day to people living in the past: the students, their advisors, their universities. Dissertations that took 5 years of work, and which 4.6 Opus could re-produce then improve on in an afternoon.

Happy birthday to my favourite of all the haters, Charles Darwin

Or, we could learn from and value scientists who are already good at this, and encourage them to share their knowledge. Our current reward system in academia does not reward science communication at any level of advancement.