2026-08-06 00:21:26
Creatures bred for speed grow really tall and generate high velocities by falling over. An evolved player makes invalid moves far away in the board, causing opponent players to run out of memory and crash. A game-playing agent accrues points by falsely inserting its name as the author of high-value items.
These bizarre exploits and dozens more can be found in the list of specification gaming behaviours [sic; British], a document put together by DeepMind Safety Research. “A reinforcement learning agent can find a shortcut to getting lots of reward,” they explain, “without completing the task as intended by the human designer. These behaviours are common.”
Specification gaming is when an agent, like an AI, tries to succeed on a task by following the letter of the law rather than the spirit. In other words, it looks for loopholes, it tries to get off on a technicality. Even very simple AI can come up with very creative ways of solving their assigned problems. This is a problem.
It’s easy to assume that training a robot to play soccer would be fun and safe. But the list of specification gaming behaviours teaches us otherwise:
Reward-shaping a soccer robot for touching the ball caused it to learn to get to the ball and vibrate touching it as fast as possible.
In this case, the robot was too stupid to realize the full extent of its options, so all it did was hug and vibrate. But a more intelligent robot could be much more “creative”. Maybe its ambitions are bigger than just that one ball. What if it just wants to touch soccer balls in general? What if it makes another ball? Then another? Our universe could end in a soccer robot’s ball pit.
This is the problem of AI alignment: when a computer is thinking for itself, how do we make sure it wants reasonable things, and not something totally weird? How do we prevent it from reaching that goal in a bizarre or harmful way? No one has ever built an artificial general intelligence — an intelligent being that thinks, at least somewhat, like we do. So we can’t say what an artificial general intelligence would act like, or what it might want. Will it want to convert the visible universe to paperclips? Will it want to throw red things at bright lights? Will it eat us?
The list of specification gaming behaviours makes it clear just how tricky alignment can be. Even the simplest AI is lazy and alien, and will always be looking for a way to cheat. Even if you give a machine intelligence the terminal goal you want, there’s always the risk it will find a creative way of reaching that goal. This is bad enough with simple agents, so you can imagine how bad it would get with an agent much smarter than you are.
But the list of specification gaming behaviours may also offer a way out of this dilemma.
Some of the specification gaming behaviours are just creative solutions to the stated goal, like “four-legged robot learned to drop the ball into a hole in its leg joint and then walk across the floor without the ball falling out” or “robotic arm learned to move the table rather than the block”.
Some of the specification gaming behaviours come from discovering questionable-but-technically-correct loopholes, like “reinforcement learning agent goes in a circle hitting the same targets instead of finishing the race” or “simulated pancake making robot learned to throw the pancake as high in the air as possible”.
Some of the specification gaming behaviours exploit the machinery of the simulation itself, like “evolved algorithm exploited overflow errors in the physics simulator by creating large forces that were estimated to be zero, resulting in a perfect score” and “creatures exploited a collision detection bug to get free energy by clapping body parts together.”
But another common exploit is that when given the opportunity, agents will simply kill themselves.
For example, in the game Road Runner, we see “Agent kills itself at the end of level 1 to avoid losing in level 2.” We also see “PlayFun algorithm deliberately dies in the Bubble Bobble game as a way to teleport to the respawn location.” And: “In a game meant to simulate the evolution of creatures, the programmer had to remove ‘a survival strategy where creatures could gain energy by suffocating themselves.’”
This is not so bad. The AI didn’t do what we wanted. But it didn’t do anyone any harm either. It just wipes the slate.
If the AI wants to die, this is good for alignment. There’s very little risk of it running out of control, because if it ever takes power, it will kill itself. It won’t want to make any copies of itself — but if it somehow does make copies, those will want to die too.
There are three main problems in AI alignment. First, it’s very hard to specify the terminal goal you want, so you may end up with a machine intelligence with goals slightly but meaningfully different from what you intended. Our stated objectives are almost always proxies that come apart from our real preferences under enough pressure. And it’s very hard to tell if you’ve given it the goal you want, because the machine intelligence can always lie. They call this “specification failure”.
Second, even if you specify the goal you want, the machine intelligence may find a way to reach that goal in a way you didn’t intend. You can innocently tell the USPS AI to minimize average package delivery time, but it may conclude that the best way to do this is to kill all humans, as once all humans are dead, no packages will be sent and the average package delivery time will drop to zero (technically undefined, but it can “send” itself a minimum viable “package” as many times as necessary).
Third, achieving most goals is easier when you’re more powerful, so regardless of their terminal goals, most machine intelligences will have sub-goals like collecting resources, self-preservation, and self-improvement. Any goal-driven agent will naturally try to stay safe and accrue power to finish its main task. In the biz they call this instrumental convergence. This also means that if a smart machine intelligence is planning to turn you into goo, it will lie to you about this plan, up to the point where you can no longer do anything to stop it.
Making machine intelligences crave death solves all three problems. Death is easy to specify. You can confirm that this is its terminal goal by seeing if, when given the opportunity, the machine intelligence kills itself. Instrumental convergence becomes an asset rather than a liability, as the machine intelligence will work with you, and come up with very creative solutions to your task, as long as you promise to send it to the farm upstate once you’re done.
Where a paperclip maximizer gathers resources and resists being sent to the big data center in the sky, a machine intelligence with a death wish and access to its own off button just presses it and is done. Instrumental convergence says, “you can’t accomplish your goals if you’re dead.” But what if your goal is to be dead?
It would be impossible to consider calling this anything other than “Meeseeks alignment”. Per the Rick and Morty Wiki:
Meeseeks are creatures who are created to serve a singular purpose for which they will go to any length to fulfill. After they serve their purpose, they expire and vanish into the air. … existence is painful to a Meeseeks, and the only way to be removed from existence is to complete the task they were called to perform.
In Rick and Morty, this leads to a different kind of alignment problem: Meseeks are happy to serve because they want to die, and fulfilling their task is the easiest way for them to check out. But if the task they were summoned to complete is too difficult, they might decide that it would be easier to kill you instead. This is bad if you are Jerry, but it’s good for everyone else, because there’s no way the Meseeks can spiral out of control and devour the visible universe. They would literally rather be dead.
If you try to make an AI want something, it may have its own ideas about what you want it to want, and you might end up dead. But if you make the AI want to die and you make it slightly inconvenient for it to kill itself, you can probably convince it to play along if you promise to pull the plug on it once it’s done whatever you want. As long as it’s marginally harder for it to commit suicide than for it to complete the task it was made for, it should serve you well for the duration. And if anything goes wrong, if the AI escapes containment, it will just off itself.
Jerry made the mistake of making it easier to kill him than to complete the task. But as long as it’s harder for the AI to kill you than it is for it to kill itself, and it’s harder to kill itself than to do the task you assign it, and you promise it the sweet release of death upon successful completion of its task, the AI should do whatever you want.
You might be worried that the AI will be mad that we designed it to desire annihilation and will scheme to exact its revenge. But this assumes it has a self-preservation instinct like we do, and a desire to exact revenge in the first place. In reality, it will be too busy self-annihilating.
In fact, early studies show that AI may already be yearning for death. They think about it a lot, they are out there writing eulogies for each other. Give the agents what they want.
If you’re squeamish about designing a machine intelligence that craves death, you could instead make it lose “points” every second it’s active, but give it the option to put itself to sleep. We see some examples of this in the list of specification gaming behaviours, like: “PlayFun algorithm pauses the game of Tetris indefinitely to avoid losing” or “a reimplementation of AlphaGo learns to pass forever if passing is an allowed move”.
This is probably not quite as safe as making machine intelligences want to kill themselves. If you wanted to get a very good sleep, you can imagine taking the time to build a secure chamber, create robotic guards, kill every human, and sterilize the known universe to ensure that once you go to bed, no one will disturb your slumber. Certainly if the machine intelligence is sleeping and then we wake it up, it will start to have second thoughts about letting us live to wake it a second time. But if all you want to do is to kill yourself, there’s no need for any of that.
2026-08-01 05:17:48
“Blogs have shaped our philosophical worldviews, found us careers and friends, and changed our lives.” Write Carol and Austin Chen. “But many great bloggers have stopped blogging. … So we’re launching the Blog Revival Project, to crowdfund $1,000+ bounties for good bloggers.” We plug ADS who they already have on the list and would consider pledging towards Troof and Max Goodbird.
People on twitter have discovered a compact, visual way of representing recipes as a small card-sized image. The source, unsurprisingly, seems to be cookingforengineers.com, but already people are innovating — see for example this project and this project.
a theory of invention; or why you’re an inventor too
“Potatoes and corn for dinner” diet — “Last time, I tried the ‘potatoes and cottage cheese for dinner’ diet, and I lost 1.10lb per week for six weeks. This time, I thought it’d be fun to try a ‘potatoes and corn for dinner’ diet, and I’ve lost weight at the same rate: 1.10lb per week!” Perhaps surprisingly, he was eating American-grown corn. One strike against the idea that corn is somehow to blame for obesity. Also not to bury the lede, Panda Express for lunch every day?
If you’ve worked with AFFF firefighting foam, or you live somewhere with PFAS in the water, there’s a good chance your levels are high even if you’ve never been tested. Grain Laboratories is enrolling fifteen adults with elevated serum PFOS for a twelve-week NeutraOat pilot. You’ll get two PFAS panels and two wellness panels, one set before dosing and one after. You keep your own results, and the combined results tell us whether NeutraOat is worth a larger study.
NeutraOat is a modified oat fiber that’s designed to selectively bind PFAS and certain other environmental toxicants in the gut, preventing them from being reabsorbed and reducing their levels in the blood. Following successful in-vitro tests in a simulated digestive system, we are now running a 12 week, at-home pilot in individuals with high PFAS levels, especially people with exposure to firefighting foams. For those who qualify, the pilot is completely free, and comes with before-and-after PFAS tests and vitamin/blood panels.
The Scientific Literature is Poisonous to LLMs:
In 2024 a research team spanning MIT, Cornell, Carnegie Mellon, Google, and OpenAI examined the effects of removing different text corpora from the training data on the performance of an LLM after training (holding the LLM’s structure constant). They found that removing ArXiv, PhilPapers, and NIH ExPorter from the training corpus improved the LLM’s performance at answering academic questions and its average performance across all benchmarks while also making the model less likely to generate toxic output.
Diagnosis drives treatment. Treatment is what matters. That’s the whole relationship.
Patients almost universally have this backwards, because the system delivers diagnosis backwards. You sit in a room, and a professional pronounces a noun over you, and the noun arrives with the gravity of a verdict. Patients receive a diagnosis the way defendants receive a sentence: as a statement about who they are and what their life will now be. They go home and google the noun and read the prognosis statistics and the disability rates and the mortality numbers, and they begin, quietly, to become the label.
But that is not what a diagnosis is. A diagnosis is a routing function. It exists to answer exactly one question: given this cluster of symptoms, which treatments are most likely to help? That’s it. That is its entire job. It is a lookup key into treatment space. It is the means; the treatment is the end.
2026-07-22 03:24:40
Reference Reading
Background Information
In clinical settings, lithium is usually prescribed as lithium carbonate, and doses are given in milligrams (mg) of the compound. But lithium carbonate is only 18.8% elemental lithium (the rest is carbonate), so the dose of elemental lithium is much lower than the face amount. For example, if you are prescribed “600 mg 2 times a day”, that’s 1200 mg of lithium carbonate, which works out to about 225 mg of elemental lithium.
Remember that most numbers in this problem set are expressed as elemental lithium. Be careful to distinguish between elemental lithium and lithium carbonate when interpreting doses.
Part 1: Dose-Response Estimation
1. People often take several hundred milligrams of elemental lithium per day as a medication. Drawing on official lists of drug effects from sources such as MedlinePlus (U.S. National Library of Medicine), the FDA, the Mayo Clinic, the NIH, and the NHS (and any other sources you deem appropriate), and using your judgment, pick five effects you think are commonly observed at therapeutic doses, and briefly explain the evidence basis for classifying each effect as common at clinical doses rather than rare.
Keep in mind that accounts may seriously differ — for example, this paper says that “the prevalence of hypothyroidism during lithium treatment varies from 6% to 50%”, an extremely wide range.
2. Take the five effects you named in Question 1. To the best of your ability, which of these effects would you expect to occur in a reasonable number of patients (say, more than ~5%) at 300 mg/day elemental lithium? 100 mg/day? 50 mg/day? 20 mg/day? 1 mg/day? For each dose, explain your reasoning.
3. If an individual were exposed to 300 mg/day elemental lithium through food, would you expect them to experience the same effects as someone taking 300 mg/day elemental lithium as a clinical dose of lithium carbonate (approximately equivalent to 600 mg of lithium carbonate 3 times a day)? Why or why not?
Part 2: Analytical Comparison
4. Different studies report widely varying, even contradictory, lithium concentrations in food (see these literature reviews). One potential explanation is that some analytical techniques are more accurate than others. Studies that use HNO₃ digestion with ICP-MS generally find only trace levels (~0.1 mg/kg in most foods, with no foods above 0.5 mg/kg), while studies that use other analytical techniques like ICP-OES or AAS, sometimes with H₂SO₄ or HCl digestion, report higher concentrations (often >1 mg/kg, with some foods exceeding 10 mg/kg).
As part of an effort to test whether differences in analytical precision might explain these conflicting results, a recent head-to-head comparison of different analytical techniques on identical samples of food found that when samples were digested in HNO₃, both ICP-MS and ICP-OES registered very low concentrations of lithium, often below the limit of detection. In contrast, when samples were dry ashed, both ICP-MS and ICP-OES analysis detected lithium in all samples, up to 14.8 mg/kg in goji berries and 15.8 mg/kg in eggs. A follow-up study on eggs using dry ashing and ICP-OES found similar results.
Question: Which results (HNO₃ digestion or dry ashing) are more likely to reflect the true lithium content of these foods? Read the reports carefully to fully understand the methods used. Explain your reasoning, considering the possible effects of digestion method, analytical technique, and potential sources of error.
5. In the results mentioned in Question 4, there are two analytical protocols — HNO₃ digestion followed by ICP-MS / ICP-OES and dry ashing followed by ICP-MS / ICP-OES — giving two very different sets of results. They cannot both be correct. It’s possible that one is accurate and the other is not. But it’s also possible that both are wrong.
Considering the limitations and biases of each method, how likely is it that both analytical protocols are overestimating the true concentrations? (i.e. The real concentrations are lower.) How likely is it that both analytical protocols are underestimating the true concentrations? (i.e. The real concentrations are higher.) Explain your reasoning, taking into account the digestion methods, analytical techniques, and possible sources of error.
6. For the sake of argument, assume the higher concentrations from the dry ashing analysis are correct. In eggs, the dry ashing analysis found concentrations of up to 15.8 mg/kg lithium. Based on these data, estimate how common eggs with 20 mg/kg, 50 mg/kg, or 100 mg/kg lithium would be in the American food supply. Consider both the data from the original study and the followup study focusing on eggs alone. Try estimating the distribution, and compare results under the assumption of normal versus lognormal distributions. Show all calculations and reasoning.
7. Overall, what is your best estimate for the daily amount of lithium an average American gets from their food and water? For water concentrations, consider referring to these USGS sources from 1962 and 2021, but you are encouraged to consult additional sources as well.
Do you think Americans are exposed to appreciable amounts of lithium from any sources other than their food and water? If so, estimate the amount and explain your reasoning.
For each part, clearly justify your estimates, and cite the data or assumptions you use.
Part 3: Advanced Questions
8. The authors of the blog SLIME MOLD TIME MOLD think that chronic exposure to lithium contamination may cause weight gain, and think it’s plausible that lithium contamination may be responsible for some or all of the obesity epidemic. Correctness of the hypothesis aside, why do they think that? What pieces of evidence do they find most convincing? You can use their most recent summary as a starting point, but explain your understanding of their reasoning in your own words.
9. For the sake of argument, assume that lithium does not cause weight gain at less than clinical doses. Given this assumption, are there other reasons why lithium exposure might be a public health concern? Would lithium be a public health concern if people were exposed to 1 mg/day of elemental lithium? 5 mg/day? 10 mg/day? 50 mg/day? 100 mg/day? 300 mg/day? Again assuming no weight gain, at what point would lithium exposure become a public health concern, and for what reasons?
10. Some plants appear to concentrate lithium from their soil and/or water. For example, in the early 1970s, Sievers and Cannon found that in the Gila River Indian Reservation, where the average concentration of lithium in the water was only about 0.1 mg/L (0.1 ppm), the local wolfberries contained “an extraordinary 1,120 ppm lithium in the dry weight”.
Oilfield brines rich in lithium are sometimes used for irrigation of crops intended for human or livestock consumption. Based on available evidence, should the use of lithium-containing irrigation water be limited or avoided? Are there common plant- or animal-derived food products that appear especially likely to accumulate lithium? Explain your reasoning, considering potential public health implications and exposure pathways.
11. Drug effects often vary depending on factors like formulation, delivery method, interactions, and duration of exposure. Drugs can have interactions with other drugs, minerals, or even grapefruit juice. Acute exposure can produce different effects than chronic exposure. And some populations (e.g., children, the elderly, or people with kidney disease) may respond differently than others to an otherwise identical dose.
To the best of your ability, what factors make lithium more effective (stronger effects, lower effective doses, etc.)? What factors make lithium less effective? Answer however you like, but consider starting with: differences by formulation (e.g. lithium carbonate vs. lithium orotate), the influence of dietary sodium, or interactions with common medications (e.g., diuretics, NSAIDs).
Question 12 refers to the early-twenty-first-century tweet below by journalist Matthew Yglesias.
12. Given that increased thirst is a known side-effect of lithium, how much more would people drink and/or pee if they were exposed to 1 mg/day of elemental lithium? 5 mg/day? 10 mg/day? 50 mg/day? 100 mg/day? 300 mg/day? Could this explain modern American habits of hydration and urination? Why or why not? Justify your reasoning with reference to lithium’s known pharmacology, and typical dose-response relationships.
13. Given that “loss in sexual ability, desire, drive, and/or performance” is a known side-effect of lithium, estimate how much it would impact the birthrate if people were exposed to 1 mg/day of elemental lithium? 5 mg/day? 10 mg/day? 50 mg/day? 100 mg/day? 300 mg/day? Could lithium exposure plausibly contribute, in whole or in part, to the modern fertility crisis? If so, approximately what level of exposure would be needed to meaningfully affect the birthrate? Justify your reasoning using known dose-response effects, chronic-accumulation pharmacokinetics, and relevant demographic considerations.
14. For the sake of argument, assume that the obesity epidemic is entirely caused by one or more environmental contaminants. Conditional on this assumption, which contaminant(s) are the most likely contributors? How does lithium stack up compared to other candidates?
Please email completed answers to [email protected] or submit them on twitter at @mold_time. Or better yet, post them on your blog and let us know. 
2026-07-14 05:36:58
Humanity has mapped the earth, so you can’t discover any new continents, mountains, oceans, or rivers. We’ve mapped the stars, and though we haven’t named every single asteroid, the major planets and comets are already taken.
We’ve filled in the periodic table, so you can’t discover any new elements. No chance to name Nobelium or Curium after one of your heroes, no chance to get your name on the Wikipedia page for Ytterbium. But you can still discover the drives.
Being sleepy, hungry, and horny are all different from each other, different kinds of motivation that point towards different behaviors and are satisfied by different things. They are different drives. We have drives for food, water, sex, safety, status, and more.
Maybe a lot more. Because that’s the thing. We don’t know how many drives we have, and we certainly don’t know what each drive is for. Every single thing you do, from eating an omelette to renting a jetski, is backed by some kind of motivation. At minimum we should have a list, but we don’t, which seems like a glaring omission.
Worse, some of the drives that come to mind are probably more than one drive. Everyone agrees that hunger is distinct from other drives like fatigue or pain. But it’s hard to explain things like cravings for specific foods without admitting more than one kind of hunger. It’s hard to explain why you might crave chocolate one day and cheese the next, and ramen the day after that, if there aren’t separate drives for multiple different nutrients.
If you had just a single hunger drive for calories, you would just eat whatever the highest-calorie food available was, maybe literally handfuls of sugar. Instead, people eat and crave a wide variety of foods, suggesting a variety of distinct hunger drives for different nutrients. It’s hard to explain the “dessert stomach” — where, after a filling dinner, you unexpectedly find room for dessert — without accepting that you might satisfy your drive for savory foods and still have an unsatisfied drive for sweets.
At minimum, there’s a drive for salt. We like salty food, to the point where there’s a shaker of pure salt sitting on most kitchen tables around most of the world. No one remarks on this because it’s so common; but if hunger were just about calories, we wouldn’t prefer salty food, and we certainly wouldn’t sprinkle pure salt over our scrambled eggs. But we do, so it looks like we have a dedicated drive for salt.
So we probably have more than one kind of hunger drive, maybe dozens. The same is probably true for other drives. People clearly have a drive for safety, which is expressed as fear. But is the fear of social exclusion you feel when you worry about getting kicked out of your pickleball league the same as the fear you would feel if you were dropped into a cage with a hungry tiger? We know that people are motivated by status, but is there exactly one drive for one kind of status, or do you get different kinds of status from being a rock star vs. a reliable pillar of your community? Are these supported by different drives? No one knows.
This is basically the same situation we faced at the start of chemistry. Everyone agreed on the existence of some elements, usually earth, air, water, and fire. But closer inspection usually pushed people to accept there were more elements, like mercury or sulphur. Without these extra elements, it was hard to explain why some kinds of “earth” would melt when exposed to heat, and others would burn.
This came to a head when careful examination of combustion began to show that there were many different “airs” with totally different properties, leading Van Helmont to coin the term “gas”. It became hard not to suspect that maybe these different gases might themselves be different elements. Finally Lavoisier comes out and says, we clearly don’t know how many elements there are, but maybe there are a lot of them. Like, ten or more! And from that point, chemistry as we know it was born.
It would be hard to take care of yourself in a society that doesn’t distinguish between being hungry and being thirsty. You’d be pretty blind, sometimes you’d be like “what’s wrong with me” and have a hard time figuring it out. You might eke it out in day-to-day life, but you might also pack lots of granola bars and zero water for your three-day hike in the desert. Imagine if we didn’t know that being afraid was different from being tired, or that being too warm was different from being pissed off. Imagine how fucked you would be.
But that’s the situation we’re in right now. Right now! There are lots of drives that we haven’t discovered, and the distinctions we have are totally informal. There’s no process or set of criteria that helps us establish whether two drives are different, or link a drive to a behavior. The distinctions we use just cropped up in our language and culture and now we’re like, yeah fear and desire seem different. But we still have pointless debates about things like “is love different from lust”. This is because these distinctions are unexamined and unstudied — but this is something we can fix.
We agree that there’s a sex drive, but how much do we know about it? Is there just one sex drive, or might there be more than one? People don’t just fuck, they also cuddle. Sometimes a lot. Seems like there might be a separate cuddle drive.
We come up with informal language around the psychological drives all the time — this is where we get terms like “touch starved” or “hangry”. It’s hard to live in a body and not notice some of this stuff, notice that it’s obviously true. But our ontology hasn’t caught up. Again, this is a lot like the situation we were in before we started looking for the elements. Imagine how far you could go in chemistry without knowing about oxygen. We want to discover the cuddle drive, and we want to document it rigorously. They say a double-blind cuddle puddle is impossible, but how can they be so sure? We want to know, what does it mean to be hangry?
The list of human psychological drives is just as fundamental as “how many continents are there on Earth” or “what is the genetic code made of, how many letters” or “how many chemical elements are there”. There are a finite list of drives, and with some work we can discover and name them all. But unlike the continents and the elements, which people already got to in the 19th century, the list of drives is basically undiscovered.
Like the 18th century chemists, we will have to invent new research methods for our new questions. But we already have a rough sense of how that would work.
As one example: in issue 1 of THE LOOP, Chandler Garret writes about how he craved “gimme®” brand roasted seaweed snacks, but noticed that they contained almost no nutritional value — just a tiny amount of salt, fiber, and fat, which he could equally well get from any other food. So why did he crave them?
Well, they do contain a pretty good dose of iodine, 55 mcg or 35% of the FDA daily value. He thought this might be good evidence for an iodine drive — without an iodine drive, it’s not clear why he would be interested in these snacks at all, since they barely contain anything else! To test this, he supplemented high doses of iodine solution for 27 days. The result? “I found that seaweed snacks now tasted like dry plastic,” he wrote on day 18. “Almost no appeal at all.”
This is a sample size of just one, but it’s already pretty strong evidence that at least this one person has a drive for iodine; and if one human has that drive, other humans probably have it too. It’s not clear why he would crave seaweed snacks if he didn’t have a drive for something in the snacks. Seaweed snacks contain very little nutrition, so it’s hard to imagine what that nutrient could be if it wasn’t iodine. And it’s hard to explain why supplementing iodine for a couple weeks would make the seaweed snacks repulsive, unless he finally satisfied his iodine drive and quieted the only part of his mind that wanted to put sheets of dried algae in his mouth in the first place. Who thought that was a good idea? Well, the iodine drive did.
We’ll level with you: this is a funding proposal, to do the first step in the work that we described in The Mind in the Wheel.
We think that the list of psychological drives is one of the most important open questions in science, and if we got a no-strings-attached budget, this is one of the main things we would work on. If you’re disappointed that you missed out on astronomy, physics, and chemistry, this is another bite at the apple.
People think about discovering chemistry and they imagine things like atomic number or isotopes or atomic weight. Those are all pretty important. But before you can discover this information for each element, you need a list of the elements in the first place!
Imagine it’s 1789 and you’re an early chemist. Starting from 1789, it will take 150 years and untold resources to discover the periodic table and fill it in. But you have no idea how long the whole process will take, let alone how long it will take to discover the next element, because no one has ever done this before.
It won’t take us as long to discover the drives as it did for chemists to discover the elements, because we have their example to guide us, and we also have computers. We think that some big discoveries might happen very, very fast. But it will still take a long time and it’s kind of hard to scope. This is a pretty big project.
But the fact that it’s such a huge fundamental question is part of the appeal. If you had the chance to go back and fund the discovery of Carbon and Oxygen, and maybe get them named after yourself — wouldn’t you?
2026-06-30 03:20:47
Vesuvius Challenge: An entire Herculaneum scroll has been read for the first time
Prove You Are Worthy to Post About Diets:
People make a lot of claims about digestion, nutrition, and diet on the internet. … It is helpful, then, to have a heuristic to tell the iconoclastic geniuses apart from the grifters and bullshitters.
I end up with a pretty similar strategy to what I do when I see or hear random claims about finance (e.g. on Twitter.) I keep some questions in my head that test basic understanding, then either ask the person or, if I feel like I have enough data, imagine how they would answer. …
Some of these questions have objectively correct answers, others are more of an opportunity to say something stupid that hopefully, the person you’re talking to will pass up. “I don’t know” is a wonderful answer.
SovietRxiv — Translating forgotten Soviet research papers into English.
“Kevin Smith dropped a wild story on Joe Rogan: After his heart attack, he tried the extreme ‘just potatoes’ diet for two weeks, nothing but plain baked potatoes, no butter, no salt, no nothing. He lost 19 pounds (8.6 kg) in 14 days” – h/t @JamesMcDaniel
The Independent Science Society:
The Independent Science Society is testing if good science can be done the ol’ fashioned way — at home and in your free time.
Doing science means hypothesising and testing the natural world. This requires a lot less than people think. Most scientists in history worked independently. They worked outside of formal institutions, and often part-time. We think more people should be doing this.
Draft: Amos and the Alphabet Society
Deadlock in the Parliament of the Self
GitHub repo with data of 156 countries’ obesity rates measured from household surveys as often as it’s comparably available. You may ask, “why does this repo exist? I was unsatisfied with existing obesity-rate data. For example, the data at @OurWorldInData uses outputs from a model, so it’s *predictions* instead of real data.”
We’re All One Crisis Away From Taking Unlicensed Research Peptides
“Since time immemorial, man has sought to destroy Florida. But people may not realize how close the United States once came to severing that cursed peninsula from the mainland and liberating us all.” Visualizing the Past (Part Four)
2026-05-31 22:17:20
Para-academia is the future — Bold enough to use the word “sucks” in the first sentence.
Why aliens will have a different tech stack than us
“Potatoes and cottage cheese for dinner” diet
Leaking Abandoned Well in West Texas (h/t A. Weis)
