José Luis Ricón
Chase every grand narrative back to its primary sources until it falls apart, then run into the knowledge that can't be written down

Glossary

A strange thing: a ledger of paid-for mistakes

Nintil has a page called "Mistakes". It is a table. Each row lists the date, the post, what was wrong, how much he paid, and who pointed it out.

In September 2019 he had claimed that neutrino oscillations have no practical application; someone corrected him and he paid $2. In 2020 Laura Deming emailed him about several errors in the early history of C. elegans research in his 20,000-word Longevity FAQ. By his own price list that came to less than $60, but he "decided to make a special exception," paid $60 and donated it to the anti-aging research foundation SENS. In 2022 he had written, in a post on AI risk, that chimpanzees have better working memory than humans, on the strength of a single study, when a review from earlier that year said otherwise. He paid $64. One row's reward column reads: "A drink next time we meet lol".

The price list is public. A minor argument that doesn't affect the conclusion: $4. A major point that invalidates a section: $20. The whole article is wrong: $200. A cited paper that had already been retracted or failed to replicate before he published: $60. In late 2019 he doubled the rewards. He modeled the scheme on the computer scientist Donald Knuth, who mails checks to readers who find errors in his books. The bounty page opens with this line:

Nintil aims to be the world's best blog as per my own criteria of what the best blog is, and one of my requirements is correctnessProve me wrong, earn money!, 2019-08-06 (updated 2026-09-08)

A blogger who pays strangers to find his mistakes tells you a lot about himself. But the most interesting part of the ledger is its latest rows. In September 2026, six entries arrived in a single month, all aimed at his recent long posts on measuring aging, systems biology and Alzheimer's. On the same day as most of them, he added a note to the top of the bounty page: the program was paused, because "I recently left my job" and he wanted to have as few obligations as possible.

Who he is

José Luis Ricón is Spanish. His mother lived in the Canary Islands, never went to college, and stayed home to raise him and his two brothers. He taught himself to code, worked as a software engineer, was at Twitter in 2019, and lived in London for a time. He describes himself as a "non-PhD holding non-biologist" who nonetheless ended up reporting directly to the CEO of the longevity company Retro Biosciences. The first time he visited Retro he went to ask for advice on starting a company. Its labs were built out of shipping containers inside an old waffle warehouse, with the HVAC system designed by the CEO. After seeing the warehouse he decided to join them instead. He left that job in September 2026. He is still blogging.

The oldest posts on the blog are 2007 notes on installing Ubuntu, then two short stories in Spanish from late 2010: one about a crime scene where a wristwatch has been stopped at 4:44, one about a secret project in a bunker under a Nazi rocket base in 1945. In 2013, in his early twenties, he wrote an "Engineerist Manifesto" parodying the Communist Manifesto ("A specter is haunting Europe: the specter of engineering"). It proposes a "Supreme Engineerium" to run the world, abolishes summer, and requires ordinary people to pass a rationality test before voicing an opinion in public. The post was tagged "humor". Commenters still took it seriously.

What he keeps wrestling with

On his Start Here page he says Nintil has no topic and that the only common thread is that he wrote it all. Then he explains where each post starts:

The motivating aims behind each of them tends to be that I want an explanation for something and nothing available is good enough.Start Here

Right after comes a sentence of almost alarming ambition: if he has been completely successful, a Nintil post represents the state of the art in its topic, "and there should be no comment possible because that would have already been thought about and incorporated."

How did he run into this problem? He has told the origin of the Soviet series himself: a video by a Soviet sympathizer, citing data from non-Soviet organizations like the CIA. That kind of argument works because most people can only pick a side between two stories. Hardly anyone goes and opens the data. He wanted to be the one who did.

Most of his early Spanish writing was political economy: the Austrian school, monetary theory, anarcho-capitalism, Rawls. He was then a committed libertarian, and rebuttals were his usual form. After he switched to English the commitments stayed, but he wrote less and less about who was right and more and more about what we actually know. The root question he hasn't let go of for over a decade: when a widely believed claim is traced all the way down, how much of it is left? And beyond that, how does human progress actually happen, and can it be made faster?

Over time the question split into two directions. One went outward: checking history, technology curves and the evidence on science funding, and eventually running a research project himself. The other went inward, late and suddenly: some knowledge, traced all the way down, turns out to bottom out not in a literature but in people's bodies and feelings. The four ideas below grew along that path, one after another.

His ideas, told as how he found them

1. Grand narratives don't survive a trip to the primary sources

In 2015 the economist Mariana Mazzucato's book The Entrepreneurial State was at the height of its influence. Its most striking example was the iPhone: Jobs's genius and design mattered, but nearly every cutting-edge technology inside it, from touchscreens, GPS and lithium batteries to Siri, traced back to government and military research money. Without that, Apple's qualities "might have led only to the invention of a new toy."

Ricón did something laborious. He found and read every bibliographic reference cited for every paragraph of that chapter, and where he couldn't get hold of one, he found other sources. He stated the method at the start: her own bibliography would be enough to criticize her own points.

One by one, the claims loosened. Before its IPO Apple took money from a "Small Business Investment Company," which the book counts as state investment. He checked a San Francisco Fed report: such firms are private venture funds that can get government guarantees, and before this one came in Apple had already raised over half a million dollars from private investors. The giant magnetoresistance effect used in hard drives was indeed discovered by two European physicists, but the people who carried it from the lab to the factory were IBM researchers who tried more than 30,000 multilayer combinations with a faster, cheaper process; the very source the book cites says so. On touchscreens, he found that Samuel Hurst invented the resistive touchscreen while on leave from a national laboratory to teach at a university, because he had a huge stack of strip-chart data to read that would have taken two graduate students about two months. Hurst later started his own company to sell it. As for cellular networks, the book cites a single report, which offers one unsupported sentence about military support for radiotelephony. He simply wrote:

Again, it is left as an exercise for the reader to try to find a link between the development of this technology and the State.The Entrepreneurial State: The case of the iPhone (II), 2015-08-21

His final point was not "government is useless." It was that what makes the iPhone the iPhone isn't the screen or the memory, which other high-end phones had too. It's the design, the operating system, the integration and the quality control, and all of that happened inside Apple. The state was one actor among many, alongside IBM, Bell Labs and Sony. The "Great State Theory" line quoted in the Essentials comes from here, and just before it he wrote:

Science and Technologies have heroes who make it go forward step by step, not Gods without which progress is impossible.Same post

A year later he did the same with the Soviet Union. His opening question: the people who say Soviet citizens ate well have data, and the people who talk about famine and queues have stories, so who is right? The FAO curve shows Soviet calorie intake consistently above the American. First he checked the FAO's method and found it was a "food balance sheet," which measures supply. Then he compared it with the USDA, an economic historian and official Soviet statistics. The numbers didn't line up: "someone is making a mistake somewhere." His interim conclusion was restrained: in terms of calories, Soviet citizens were reasonably well fed.

Then he went to the émigré. Igor Birman was born in Moscow in 1928, was director of planning at three factories, emigrated to the US in 1974, and there criticized the CIA's estimates of the Soviet economy. Ricón quotes another economist's judgment that, on getting the Soviet economy right, the most "advanced" models had done worse than Birman's "anecdotal economics." Birman's book gives food three chapters. Calories were about equal, but total value was under half the American level. Canteen meatballs were nearly all bread. Potatoes were often frost-bitten, and milk sometimes spoiled in the shop and was sold anyway. From a separate émigré interview study, Ricón adds another number: the USSR was the world's largest milk producer, yet only about 60% of its milk was consumed by humans, 36% was fed to animals, and the rest was wasted. In the US the figure was around 90%.

The series had an ending he himself found amusing: "the series has been cited by communists and anti-communists alike in support of their opposing positions." That sentence is itself a side effect of his method. When he traces something to its primary sources, what he hands back is usually not a verdict but a more detailed map: which claims hold up, which don't, and which nobody has measured. People on every side can find the patch they wanted.

2. There is no "Great Stagnation", only specific bottlenecks

In the mid-2010s a popular idea among economists was that the rich world had entered a "Great Stagnation" around 1970; Tyler Cowen and Robert Gordon each wrote a book about it. The evidence was slower growth in total factor productivity and GDP per person.

Ricón came at it as an engineer. If technology is a big part of productivity, why not look at technology directly? In 2016 he wrote No great technological stagnation and laid out every technology-progress curve he could find: information technology, photovoltaics, batteries, coal-fired boilers, gene editing. Some fields had stalled (aircraft speed), some were improving (gene editing, solar), and some were getting worse (the cost of each newly approved drug). And none of these curves had an obvious kink in 1970. The economic slowdown was real, but it was not a slowdown of technology across the board.

In 2021 he came back to it and pushed further. The trouble with the "Great Stagnation" framing, he said, is the word "great": it tempts you to look for a single cause and then fit every field into it. His answer to the question "is there a stagnation?" was a single word, the Zen koan answer "mu", which means roughly: un-ask the question. He told a story. In the year 10,000, chips at TSMC (Intel went broke in 2050) have been stuck at 1 nm for decades, and a boss yells at someone that at historical rates they should be below the Planck scale by now, so why so slow? The man replies: sir, I'm the janitor, and fundamental physics doesn't allow it.

Rampant optimism is a recipe for constant disappointment and self-flagellation.Still no great stagnation, 2021-01-27

His alternative was to set the big picture aside and do a root-cause analysis field by field. Why isn't cancer cured? Look at the problems inside oncology itself. Why no flying cars? Look at batteries, composite materials, the fluid mechanics of low-noise fans, the economics of short flights, and the regulations.

He actually did this. In early 2020 he published The Longevity FAQ, around 20,000 words, for which he says he reviewed more than 100 papers, most of them reviews themselves. It is the post his later posts link to most. It contains a very Ricón moment. Resveratrol was being sold as an anti-aging molecule, yet the US National Institute on Aging's mouse-testing program found no effect. Some said the dose was too low. He did the arithmetic: the mice got 200 mg per kilogram of body weight, which works out to 16 grams for a human, while the molecule's best-known promoter takes half a gram a day. The mice got 32 times his dose, "so clearly the issue wasn't that the dose was low." (In 2022 he added a note that he had grown more skeptical of resveratrol and NMN.)

The FAQ ends on a structural problem. Telomere researchers, mTOR researchers and sirtuin researchers each work on their own mechanism, but if these mechanisms are tightly linked, fixing any one of them alone won't show much effect. That is what he means by a bottleneck. In 2020 he left tech to study the bottlenecks in aging research full time, and the work became a concrete project, Rejuvenome: a $70 million collaboration announced by the Astera Institute and the Buck Institute to test combined interventions in large numbers of mice. Between a blog post rebutting the Great Stagnation and a room built for thousands of mice there was only one idea: don't ask why the world is stagnating; ask where this one field is stuck.

3. He checked the most tempting conclusion, "give good people money", then put his own field on probation

In 2019 an essay in The Atlantic calling for a new "science of progress" cited a striking number: the economist Pierre Azoulay and colleagues had found that scientists with long-term HHMI funding were 96% more likely to produce breakthrough work. It became the most quoted evidence for the science-policy slogan "fund people, not projects."

At the end of 2020 Ricón started a five-part series named after the slogan. In the first part he took the paper apart. The 96% came from an estimate the authors themselves acknowledge is more exposed to pre-existing trends. With the more robust difference-in-differences specification the figure was 43%: still substantial, but not 96%. The bigger problem was selection, and he explained it with fund managers. Look back and pick the 100 best-performing funds of the past five years, and you get a mix of the genuinely skilled and the lucky, indistinguishable beforehand. Afterwards, the lucky ones regress to the mean while the skilled keep rising. A control group can "look, swim, and quack like HHMI Investigator material" and still be far from it. The authors had argued that HHMI investigators were nominated by their universities rather than applying, which made it hard for risk-loving scientists to sort themselves in. He pointed out that HHMI has since switched to self-nomination. His verdict:

So does the paper live up to its promise of 96% more high impact citations? I don't think so.Fund people, not projects I: The HHMI and the NIH Director's Pioneer Award, 2020-12-27

But he wasn't out to demolish the idea. In the third part he tested the "Newton hypothesis": is science mainly driven by a small elite? Working through the citation data, his answer was yes. The elite might be 1–5% of practicing scientists. So "fund people" makes sense in principle; the open questions are whether you can pick them and whether picking them is worth what it costs. His compromise was to split by field: biology probably benefits more from funding projects, theoretical physics more from funding people.

Over two years he wrote roughly 80,000 words on metascience. At the end of 2022 he wrote a summing-up that opens by admitting something uncomfortable. He suspected that one post he had written about US immigration might be more useful than all his metascience writing combined, except his bottleneck work on aging, which had led to a real project. Reading all those books, papers and conference recordings had been like

trying to squeeze a dry lemon for not, it ends up being, much juice.Limits and Possibilities of Metascience, 2022-12-05

His argument starts from a question posed by Michael Nielsen and Kanjun Qiu in a 30,000-word essay. If aliens did science, would they have PhDs, funding agencies and peer review? Nielsen and Qiu thought probably not, since human scientific institutions depend so much on historical accident. Ricón's answer was: "Probably."

He explained with restaurants. Restaurants everywhere are much the same: a kitchen, seating, menus, a bill. That isn't because one chain has a monopoly and crushes experiments; there are millions of restaurant owners. It's because people need to eat, like to eat with friends and family, and trade off money against time, so that is what a restaurant should look like. If aliens are like us in the relevant ways, they will have restaurants too. Likewise the PhD: aliens are presumably born knowing little and later become scientists, so something has to happen in between, and as long as tacit knowledge exists, someone has to apprentice someone. "At that point we have hand waved our way to a proof of the necessity of something like the PhD system."

So he is cautious about "more institutional experiments." He cites a success, the Thiel Fellowship: no randomized trial, about $30 million over eleven years, and it helped launch people like Vitalik Buterin and Dylan Field. It worked because someone saw a specific opportunity and acted on it, as in entrepreneurship, which doesn't turn into a testable general theory. His conclusion: metascience should be "institutional entrepreneurship", with research and execution done by the same people. He predicted that academic metascience and the people who actually reform science "will continue walking separate paths." He also put his own taste on the table:

Forcefully: Do you rather want a better CRISPR or the next quantum mechanics? This is probably the crux question. I say obviously give me the next CRISPR.Limits and Possibilities of Metascience, 2022-12-05

The essay closes by saying it is his last on metascience, at least for a while. Judging by his later posts, he hasn't written another long piece on it since. Someone who spent 80,000 words studying how to make science faster concluded that the field has little to offer: pick people well, give them money and time, and then go do the work yourself. And he did go and do it.

4. The most important knowledge doesn't fit in a paper, including his own

This idea started with a picture. In autumn 2021 the research funder Ben Reinhardt posted a diagram of two ways to learn a field. The "right" way is to go around talking to people. The other, drawn as a misconception, is to read the key papers. Ricón said he was exactly the person in the picture who only reads papers, wanted to respond, and ended up writing a 13,000-word essay called Scaling tacit knowledge.

He first sorted knowledge into kinds. There is public explicit knowledge (facts you can look up) and private explicit knowledge (real reagent prices; "that paper from 30 years ago? A friend was in that lab, it's quite sketchy"). Then there is tacit knowledge, which itself comes in several kinds: motor skills (riding a bike), judgment (telling when a steak is done), and knowledge that isn't in any single head but in a whole organization (he claims the CEO of TSMC doesn't know how TSMC works).

Then he gave two unsettling examples. One is Fogbank, a material used in US nuclear weapons that hadn't been made in decades. Recreating it took many years and millions of dollars, even with the original instructions and access to the people who had worked on it. The key turned out to be an impurity that nobody on the original team knew about. The other comes from the sociologist Harry Collins. Western labs tried to replicate a Soviet measurement on sapphire samples and kept failing. It turned out the Soviet fine thread had to be greased with human skin oil, by running it across the bridge of the nose or behind the ear, and only some people's skin worked. That is how the lab notebooks came to contain entries like "Suspension 3: Fred-greased Russian thread; Suspension 12: switched from George-grease back to Fred-grease."

Most people would stop there and marvel at how deep tacit knowledge goes. Ricón refused to:

We should not look at tacit knowledge in awe of its fractal richness, being humbled by its existence. Rather we should look at tacit knowledge as a challenge to be overcomed!Scaling tacit knowledge, 2021-12-10

His lever was chess. The rules are public, machines have long played better than people, there are centuries of teaching tradition, and getting good still takes years. "100h of Magnus Carlsen talking to you wouldn't be enough to get you to play good chess." So how do players learn? Going through the research, he found that the strongest predictor of chess skill was neither tournament games nor coaching but hours of serious solitary study: puzzles of increasing difficulty, and memorizing grandmaster games. One amateur went from 1200 to 1600 in six months by doing exactly that, and felt that actually playing helped less. When you memorize a game you are forced to understand it, so you remember it in chunks rather than move by move.

From this he drew a prescription: record experts doing the real work, ideally talking as they go, build a large library of it, and make learners engage actively, guessing what comes next and what the expert will notice. He wrote that if someone wanted scientists to strap GoPros to their foreheads and record how they work, he would fund it immediately. Then comes the most important turn in the essay. It is the "I used to think that" quoted in the Essentials, followed by:

A collection of anecdotes is the kind of data we need to transmit tacit knowledge.Scaling tacit knowledge, 2021-12-10

That sentence only weighs what it does against where he came from. His whole blog was built on "don't trust anecdotes, check the data." Yet the witness he trusted most in the Soviet series was an émigré whose "anecdotal economics" had read the Soviet economy correctly. Now he wrote the principle down himself.

A few years later the idea turned in a direction he probably didn't expect. In 2015, in a long rebuttal of Scott Alexander's Meditations on Moloch, he dropped in an aside:

Not being much of a fan of emotional language occurring in this blog, I’ll make a little exceptionSlaying Alexander's Moloch, 2015-12-24

In 2025, turning 33, he listed 33 lessons. One was "Turns out emotions are real and useful and I have lots of them to feel"; another, "Just being smart doesn't go as far as you thought." At the end of April 2026 his mother died of cancer; he had flown twenty hours to the Canary Islands to be with her. In May he published There was only mom, which includes this:

Growing up I had consciously chosen to repress them all, I thought they were bad and irrational.There was only mom, 2026-05-03 (he is speaking of anger, fear and sadness)

He remembers the young José feeling proud of how well he could suppress strong emotions. In September he wrote a post on choosing a therapist titled "Therapy is like dating." One of the most agreed-on predictors of whether therapy works is how well you click with the therapist, so being randomly assigned a therapist is like being randomly assigned a partner. Discussing "evidence-based" therapy, he noted that some powerful newer methods haven't been studied yet, and that there is no need to wait for them:

But here we are patients not scientists, we are not interested in highly repeatable intersubjective knowledge.Therapy is like dating: how to know if it's working, 2026-09-17

It's tempting to read this as a rationalist's conversion. Followed along his own path, it looks more like the same line reaching its end. The tacit-knowledge essay had already said that some knowledge can only be passed on through doing, through bodies, through particular people. He just found, in the end, that the tacit knowledge hardest to trace to its source was in himself.

How he thinks

Read every one of the other side's references. He checked the Mazzucato chapter paragraph by paragraph, reference by reference, and where he couldn't get a source he said so and used another. It sounds like the basics, but most book reviewers don't do it. In 2015 he published a blog post summarizing the Handbook of the Economics of Innovation that runs to 44,000 words.

Do the arithmetic. The resveratrol dose, the HHMI effect size, Soviet calories: he rarely accepts a number without converting it into another unit to have a look. Plenty of neat claims give themselves away under a change of units.

Go field by field. "The Great Stagnation", "the Great State", "the vast design space of metascience": his first reaction to any grand framework is to break it into concrete cells. Which field, which curve, which bottleneck. He even answered a question that was too big with a Zen "mu".

State his position, then admit he can't prove it. "Give me the next CRISPR" is immediately followed by the admission that he can't prove doubling down on an existing field beats starting a new one. He tends to separate "this is evidence" from "this is my taste", so readers know which parts they can check and which they can only agree or disagree with.

Put a price on being wrong. The bounty and the ledger turn "I might be wrong" from a polite phrase into money owed. Every post carries a link at the top: "Is this article wrong?" His long essays end with a "please cite this essay as" block. He writes a blog to the standard of a paper.

Check his own assumptions, not just other people's. In 2022 he wrote Stop assuming so much, and every example is one of his own blunders. He had always assumed a person could hold only one US O-1 talent visa and never checked; when a friend asked, it turned out you can have more (he held two when he wrote the post). At Rejuvenome he nearly got the capacity of the mouse room wrong, until he wrote a simple simulation and found the plan didn't fit. His advice: set aside an hour every so often to think specifically about what you are assuming.

Eventually, treat himself as data. In his 2025 post Agencymaxxing he wrote that when he played StarCraft competitively, every action was dictated by the optimal build order: "you become a vessel through which the game plays itself." He used his own experience of learning to dance to take apart why people don't do things. He was afraid of imposing on others, so he didn't ask good dancers to lead him, until he realized that dance teachers exist for exactly that. He had always been good at taking other people's arguments apart. Later he started doing the same to himself.

Where he might be wrong

First, the claims he checks hardest tend to be ones he already disagreed with. He came out of Spanish-language libertarian circles and wrote early series like "Road to Ancaplandia" (the land of anarcho-capitalism). He has taken apart the "entrepreneurial state", taken apart Soviet myths, and written on whether war drives growth, saying he is "not particularly friendly to the view that wars got us where we are." Each of those checks may be correct, but the choice of targets is lopsided: in what I read, he rarely brings the same intensity to a popular claim that favors free markets. He himself noticed that both sides cite the Soviet series, which shows that what he delivers can be used selectively. In the Mazzucato post he also says that computers, semiconductors and GPS are left out of that discussion, and those happen to be the examples most often used to show the state's role. Bibliographic facts: in 2015 the economist Alberto Mingardi published "A Critique of Mazzucato's Entrepreneurial State" in the Cato Journal, and Ricón wrote a post about it. I did not find a named rebuttal of Ricón's series, and I have not read Mingardi's paper or Mazzucato's own replies to her critics, so I won't restate them.

Second, the restaurant argument may just be a defense of the status quo. He says scientific institutions are like restaurants: they look alike because that is how they should look. But restaurants have a market constantly weeding them out, and in the same essay he concedes that science has no such chain from institutional performance to the spread of norms, that decisions are highly centralized, and that there is less experimentation than in a market. So the sameness of today's scientific institutions could be a product of too little competition rather than proof of an optimum. He prefers CRISPR to the next quantum mechanics and says openly that this is taste. But a funding landscape set by that taste is exactly the "monoculture" Nielsen worries about.

Third, his recent writing uses exactly the evidence standard he once took apart. He wrote this one down himself. In 2020, faced with a "huge if true" number like HHMI's 96%, his first move was to check the controls and the estimation method. In 2026 he recommends an attachment therapy because "in a small study, has the largest effect sizes I've ever come across," and tells patients not to wait for replicable evidence but to trust their experience. That may well be right for therapy, and his reason (whether a therapist and patient click can't be randomized) makes sense. But readers should see that this mind uses two different rulers in two domains, and he has not yet written about how they fit together.

Fourth, as his range widens, the errors are piling up. The mistakes ledger took six entries in September 2026 alone, all on his recent long posts on measuring aging, systems biology and Alzheimer's: shingles mixed up with chickenpox, a 1975 classic experiment described wrongly. He accepted and fixed every one, which is rare. But the bounty was paused the same month, so readers no longer get paid to find errors, and whether the ledger keeps being kept this way is up to him.

Fifth, a controversy I have not read. In 2017 he wrote a set of posts on why there are few women in computer science. On his Start Here page he summarizes them as concluding it "has more to do with preferences than with discrimination or stereotypes," and adds that "this one is bound to be controversial." I have not read those posts or the rebuttals to them, and only report that they exist.

Sources

• The Entrepreneurial State: The case of the iPhone (II) (2015-08-21)A reference-by-reference check of the iPhone chapter in Mazzucato's book: hard drives, touchscreens, cellular networks, and the "Great State Theory".
• The Soviet Union: The food consumption puzzle (2016-05-11)Why Soviet calorie intake appears higher than American, where the FAO data comes from, and Birman's "equal calories, half the value".
• Fund people, not projects I: The HHMI and the NIH Director's Pioneer Award (2020-12-27)Taking apart the "96% more breakthroughs" HHMI paper: estimation method, control groups and selection.
• Scaling tacit knowledge (2021-12-10)Kinds of knowledge, Fogbank and "Fred-grease", how chess is learned, and "a collection of anecdotes is the kind of data we need".
• Limits and Possibilities of Metascience (2022-12-05)The summing-up after 80,000 words: restaurants, alien PhDs, the Thiel Fellowship, and "give me the next CRISPR".

Updates