self_made_human
amaratvaṃ prāpnuhi, athavā yatamāno mṛtyum āpnuhi
I'm a transhumanist doctor. In a better world, I wouldn't need to add that as a qualifier to plain old "doctor". It would be taken as granted for someone in the profession of saving lives.
At any rate, I intend to live forever or die trying. See you at Heat Death!
Friends:
A friend to everyone is a friend to no one.
User ID: 454
On The Studied Irrelevance of Indian Schools
A note: this essay began its life as a reply to the wider discussion about Freddie deBoer and his god-awful takes on disciplinary standards inside schools downthread. I followed the link, encountered the paywall at speed. My prose went completely off track, sprawled out of control, and I've given it standalone status.
I would be more upset about all this if I believed my future kids stood to benefit from schooling. I cannot rule that out entirely, but I do not intend to start a college fund on their behalf unless the next five to ten years deviate sharply from my expectations. That isn't the same thing as not saving money on their behalf, which I absolutely do. Remember, money is fungible, and can be exchanged for goods and services.
I do not live in the States either, though some of these concerns map onto the British system, as dim as my understanding of that system happens to be.
My experience with the Indian education system, which I have more familiarity with than I would have wished, has left me with what I'd call a hot-take if most people back home wouldn't tacitly or vocally agree with me.
Bluntly: the schools you attend before college did not really matter. A prestigious name-brand private school could be expected to give you a marginally better education, but most parents who choose them are buying prestige and peer composition. The middle class and above does not send its kids to a government-run school if they can help it. The students at government schools are not particularly awful or ill-behaved, and we don't have the disciplinary problems Western schools deal with. Government schools are grossly under-resourced, and they serve a population whose parents cannot afford the parallel education system. Your kids would come out of it with an inferior social network, worse English, and a sense of shame. It's almost unheard of to pick one if you can afford better, even if that accordance requires severe tightening of physical or financial belts.
Why? The real answer is that nobody expects the teachers to actually teach you much in class. Including the teachers themselves. I am being serious about that. Even the best teachers at the prestigious school I attended knew, with the placid certainty of a thermodynamic law, that their star pupils and everyone else would be attending private coaching sessions outside school hours.
That's what you have to understand to make sense of anything that follows. School is the credentialing and networking layer. Coaching is the education layer. The two have been functionally decoupled for decades, and the entirety of Indian society organizes itself around the assumption.
You want numbers? I've got numbers. India's private coaching industry was about Rs 58,000 crore (about $7 billion) in revenue as of 2022, with projections to hit Rs 1.3 lakh crore (about $16 billion) by 2028. That's a parallel education system roughly the scale of a respectable national economy. In a country with around 250 million students in K-12, comprising almost 25% of the world's school-going children, that represents a lot of household income diverted from other purposes. Somewhere in the ballpark of 7.5 million students are enrolled in formal coaching at any given moment, with Kota alone drawing more than 200,000 annually.
The terminus, the thing all this preparation is for, is a small handful of murderously selective exams. 1.4 million students sit JEE Main each year for entrance to engineering colleges, of whom roughly 250,000 qualify for JEE Advanced, of whom around 17,000 actually get into one of the 23 Indian Institutes of Technology. The Computer Science seats at IIT Bombay close out around rank 60-70 in the country. Roughly one in a hundred thousand students gets a CSE seat at IIT Bombay. NEET, the medical-school entrance, runs on comparable arithmetic: over 2.3 million students sit the test annually for around 110,000 MBBS seats, of which fewer than 60,000 are at the more affordable government colleges. Harvard's acceptance rates are only slightly lower than the odds of getting any medical seats in India.
This is what your kid is competing for, and this is why your kid is in coaching from age fourteen, or twelve, or ten, or, occasionally, five.
I have my own theory about why parental coaching enrollment ages keep dropping. It's not a particularly original one, since it reduces to the logic of any arms race. If the median competitor starts at twelve, the marginal advantage of starting at ten is large. Once enough people start at ten, the advantage shifts to eight. There is no obvious floor, until we've reached toddlers still crawling there. ThePrint recently profiled a five-year-old in Jaipur who attends two hours of tuition after a five-hour school day, carrying eleven books in his bag. His mother has already decided he will be a doctor. He might struggle to count to thirty. I strongly suspect that we won't be recruiting doctors by the time he turns twenty.
Back to the school itself. Why don't the teachers care?
Several reasons, layered on top of each other, mutually reinforcing.
The first is selection. A physics or biology teacher actually good at teaching physics or biology can run their own coaching center. This pays better by a significant multiple and selects for higher-functioning students who actually want to learn. The teachers who stay in regular schools are typically the ones who cannot, or will not, make that jump. A few of my schoolteachers moonlighted as coaches on weekends, and the fact that they could not sustain it full-time tells you where they stood in the talent pyramid. The very best teachers I encountered were either independent coaches or worked at the apex coaching institutes.
The second reason is incentive. Schoolteachers are paid the same whether their students learn or do not. A coaching center operates on something closer to a results-based payoff structure. Parents pay a coaching center for performance on entrance exams, and they will pull their kid out and switch institutes if results are bad. It's much harder for the school to lose your tuition this way, because admission to a desirable school is itself a scarce good. Parents pay for the brand and for the peers. They do not pay for instruction. They pay out the nose for that elsewhere.
The third reason is information. Everyone in the system knows everyone else knows. The teachers know the students are getting taught elsewhere. The students know the teachers know. The parents know everyone knows. School ends up being an attendance requirement, a social environment, and a credentialing function. Nobody with their a functional frontal lobe mistakes it for the place where the learning happens.
Imagine this scenario: you're an Indian parent, and dear Ramesh (I had to pick a maximally stereotypical name, because my skin color allows me to get away with it) can't solve the Riemann hypothesis. He can't even prove the transcendental nature of pi! You accost his school teacher, and harangue him on the deficiency of his didactic approach. He is genuinely shocked. “But madam, I gave him the homework exercises. I told him it'll come on the test. I scribbled hastily on the blackboard and rapped knuckles, including his. I made sure nobody was snoring in class, or using their phone. What more can you expect of me?” You deflate, and feel suitably chagrined. You are overcome with shame, you grab Ramesh by the wrist and drag him off to whatever sprawling network of coaching centers is in vogue today. I haven't quite kept up.
The end result is a level of self-aware double-bookkeeping that I do not think most Western observers fully grasp. When you ask an Indian student "did you learn this in school?" they will laugh, albeit internally. The real answer is that they learned it in their evening coaching, with a private tutor, or by self-studying their books.
The coaching layer functioned as its own form of tracking. The smarter, harder-working students would study on their own or get more intensive and dedicated tutoring. The smart but constitutionally lazy ones, like me, attended too, probably with worse results. My ADHD was at the time unacknowledged and unmedicated, which I now realize accounts for considerably more of my misery than I appreciated at the time. Lucky for me, my parents ensured I had 1:1 private tutoring for most subjects for most of my life. It's hard(er) to daydream or procrastinate when someone is watching you like a hawk for two consecutive hours. Thanks, mom and dad? I suppose it worked.
There was also more explicit tracking, which I expect that American educators would find appalling. The duller kids got diverted into the humanities by 8th or 10th grade. Kids from mercantile backgrounds who wanted to stay in the family business opted into commerce. The would-be engineers, doctors, and professionals took the science track, with some discretion over whether to include biology, programming, and so on. Maths, physics, and chemistry were core and immutable. English literature and a native language rounded things out. I'd like to claim I did History and Geography until 12th grade, but it has been long enough that I genuinely cannot remember. On reflection, the answer is probably yes.
I have younger cousins, and I vicariously observed a rather interesting trend through them. An increasing fraction of Indian parents have recalibrated on the importance of sending their kids to the “best” schools, or at least have decided that the return on investment isn't worth it. So it went with them. They moved, around 10th grade, to a “worse” school. Why? Because that institute had subtly advertised that it did not give a hoot about in-person attendance metrics or regular grades. The understanding was that the students enrolled only as a legal formality, they were expected, even encouraged to not show up to class and instead spend that time studying on their own or attending private coaching. A remarkable innovation, really. I wish it had been around in my time. I'm surprised it took so long to show up, that's what every incentive leant towards.
After high school: well, bud, better give the college entrance exams your best shot. Your future hinges on them. For the sufficiently wealthy, paying out to put your kid in a private institution was an option, though it came with a degree of stigma. Everyone knew that the kid who got into a government-run engineering college like an IIT had more intellectual horsepower than the best private alternatives. I will set aside the affirmative action issue, partly because I do not want this essay to be about that, and partly because anything I say about it will infuriate someone. Same went for medical schools, law schools, business schools, the entire prestige hierarchy. Government meritocratic top-tier sat at the apex, followed by a long tail of private institutions of variable quality, some genuinely excellent and some basically diploma factories.
The dimmer kids? Too bad. Barring their parents, nobody expected much of them. Did they learn anything in class? Who cares. Did they learn anything from the extra coaching? If the answer was no, that was on them.
I should admit I do not have great visibility into what happens to the genuinely subpar in academic terms. Even the dullest tools in my school shed came from reasonable wealth. They probably had a sinecure waiting for them in a family business, a comfortable middle-management slot somewhere, or some equivalent arrangement. The default outcome for those not quite so lucky is probably a dead-end job and a long slog. You cannot expect better from a poorer nation. They manage. They have to, because there are hundreds of millions of them.
Disciplinary measures were far stricter than in the West. Expulsion for unruly behavior was a rare but real possibility. If you were a bad student, you had better be a well-behaved one while lessons were happening. I’m just about old enough to remember going from corporal punishment being the norm to mostly frowned upon. Mostly. The legality of it was never the primary concern for anyone involved, including the parents who knew it was happening and either silently approved or tolerated it as the cost of having their kid in a serious institution.
Reading the Freddie deBoer discourse from a distance, I am consumed with cynicism, mostly because I perceive it as charitably misplaced idealism. The Western debate around discipline often proceeds as though the alternatives are "strict enforcement with consequences" versus "permissive chaos," and we are arguing about which moral failure to commit. After serving time inside the Indian system, I observed that strict discipline turns out to be the easy part. It isn't hard to produce a quiet, orderly classroom through any number of methods, some humane and some less so. The hard question is what happens inside that quiet, orderly classroom. The answer in India is: surprisingly little, by design, or by tacit acceptance.
Do I think this is a good thing, something to advocate for? Oh god no. I find it depressing. I'd like to claim that better schooling inside the schools would help, but surely you've understood that this is another Red Queen race right? You run as fast as you can just to stay in place, and making everyone faster doesn't change the composition of those making a podium finish.
Kota, the city in Rajasthan that has long been India's largest coaching hub, makes the Indian news with depressing regularity because of student suicide. A handful a month, from memory. There would be more, if they weren't incredibly strict about preventing them. If you imagine they achieve this by improving living standards or reducing the load? There's no hope for a quokka like you. They just make it harder to kill yourself, including through the kind of anti-suicide measures usually seen only in in-patient psychiatric wards, combined with draconian surveillance. The students are teenagers, sent away from home to live in hostels and crammers, with the expectation that they will spend two years of their lives doing nothing but preparing for one exam. Most of them will fail. A small number will pass and go on to lives of unevenly distributed satisfaction. The wager their parents made on them, financial and emotional, will turn out to have been a bad one in the median case. They were never going to be in the top one percent. Almost nobody is. The maths doesn't math.
If I had to place this in wider context: imagine something halfway between the laissez-faire approach of the West and the all-consuming grind of East Asia. Thankfully it was not quite that bad. I am not fond of the system at all. It almost chewed me up and didn't quite spit me out again, and the only affirmative defense I can make of it is that the chewing has not stopped since I left, despite my having moved far, far away. It is just not as bad as it could be, which I say while noting that explaining why sacrifices were made to Moloch is a different exercise from condoning the practice or wishing it to perpetuate into the future. We burn our children to prepare them for the flames ahead. The fires do not really cool off once you are out the other side. The forms of suffering reorganize themselves around your changing life circumstances, and that is roughly the most many of us can hope for.
What I can say is that I had some slack. I had opportunities to be a child rather than meat slurry waiting to be packed into a mold. The front-loading of suffering arguably bought me optionality for the future, though I am no longer fully confident this is true. Well, it's the thought that counts.
As someone who was on the local equivalent of a pre-med track, I might have spent between two and four hours after a full school day with private tutors, but the real pressure ramped only up after 10th grade. I can't blame my parents for that either. They weren't tigers. They were on the slightly upper end of strict by upper-middle-class Indian standards, largely because they had gone through the whole ordeal themselves and knew the value of higher education in a country where higher education is the one ticket out of a great many situations. I would do the same with my kids if I had to. I do not expect to have to. If I have kids, I hope they grow up somewhere the pressure is not nearly as bad.
Then again, it is hardly that much better in the West. An aspiring American doctor better start working on extracurriculars well before finishing high school. In the US, even more years of their limited life get spent in pre-med coursework, after which they apply to medical schools that admit a small fraction of applicants and that publicly weight non-academic considerations heavily. All you have done is trade the legible, harder-to-fake signal of excellent grades for excellent grades plus sob stories about digging wells in Africa, plus the presidency of the Underwater Basket Weaving Society, with an added interest in supporting weavers from under-represented minorities, which should help look past the uncomfortable fact that you yourself are not an under-represented minority.
The Indian system is more legible and crueler about it; the American system is less legible and crueler about hiding it. I'd have to be a touch daft to claim that the former is superior to the latter, I know which one I'd choose for my kids, warts and all.
It could be worse! South Korea has roughly 75% of students enrolled in around 100,000 cram schools nationwide, and Japan has over 65% of ninth-graders attending juku. India's coaching enrollment is enormous in absolute terms but smaller as a share of total students, largely because most Indian families simply can't afford the fees. Our middle class might out mass the entirety of the United States (albeit not in a literal weighing setup; gotta wait for the Ozempic to kick in to make that a fair fight), but there are a lot of poor people here. They never had a chance.
There are games where the winning move is not to play. Those Indians should consider themselves lucky that they can't afford to play them, at such a scale. Somehow, I doubt they'd feel comforted if I were to tell them that. In fact, I strongly suspect that they'd get angry at me.
At the end of the day, schools (or education in general) are a selection device for a tiny number of slots in professions that pay disproportionately well in societies where most professions do not pay nearly as well. As long as those differentials hold, you can expect parents to pour everything they have into making sure their kids end up on the right side of the divide.
Good thing all of it is about to be moot anyway. I am writing this in 2026, after all, and you know what I expect of the near future. I will not belabor the point. Anyone reading this knows what I am referring to. If you don't, well, you have other things to worry about than your kid's MCAT score, and this essay is not the place to acquaint you with them.
God. It sucks to grow up and realize that your parents, despite their flaws, loved you and wanted what was best for you. That the pressure they put you through, or the pain they expected you to endure, was for your own good. Positive expected value, at minimum.
It would be easy to rail at my parents. I have, on bad days, indulged the temptation. I would have done the same in their shoes, however, and I would have hoped my kids forgave me for it later. The forgiveness comes easy now, long after I have passed the stage of needing my parents to shove me out of the nest, and discovered the entirely different qualia of suffering that is charting your own path while holding yourself accountable. I am told that some people find this easy. Good for them. I will settle for prescription stimulants and a regularly-examined internalization of parental expectations, which turn out to be remarkably portable and follow me wherever I go, whether I want them to or not.
I do not much like that fraud, Freud, but this is classic super-ego stuff. Where’s my suspiciously penis-shaped cigar? I could use a smoke after letting all of that out. Or, given the times, at least a good puff on a vape.
Thanks. Glad to be depressed together, and hoping we can get well together. At least you know my advice comes from a place of unusual professional and personal familiarity.
I suspect that I'm having a depressive relapse after a month or two of genuine euthymia. No surprise that coincides with a return to work and exam grind. My workload is probably 3-10 times what it was on my first rotation. I used to get bored during my shifts. Now I barely have a moment to park my ass, and the other doctors and I have to draw straws to decide who gets to have lunch first. I used to have the time to (at least in theory) revise my notes during working hours or browse the internet. Right now my phone is helpfully noting that I've reduced my screen time by a remarkable margin.
Is this going to get better? Hah. Haha. Hahahaha. This is going to be my life for the next 5 months, no relief from the pain. I am under-medicated for my ADHD. I have worsening migraines. I leave work wanting nothing more than to crawl into bed and stay there.
What clinched the self-diagnosis was sighing. Literal, audible sighing. The last time I was properly depressed, a junior colleague clocked this tell before I did, which surprised me, since I thought I was hiding things well. The literature, predictably, confirms a correlation. (The body keeps the score and occasionally narrates it out loud.) I caught myself doing it yesterday. Then an intern asked me why I was sighing.
@ToaKraka was kind enough to link to the PHQ-9 screening questionnaire lower in the thread. One glance at it made me wince, I didn't have to add up the numbers to know it didn't look good for me.
The good news is that Paper B pressure lifts next week, possibly forever if I pass. I have never failed an exam in my life. That fact is a load-bearing pillar of my self-esteem, and I am aware of how that sentence sounds coming from a psychiatry resident. I'm willing to risk the burnout. The exam has to be cleared eventually, deferring wouldn't buy me study time anyway, and a pass earns me twelve to eighteen months of academic reprieve. The workload stays the same. This is the only consideration keeping me from filing the current monomaniacal focus under "obviously irrational."
Apparently, around 20% of psychiatry residents experience burnout or depression. Lovely. Glad to have good company. I know the pharmacological management of depression like the back of my hand.
Before anyone panics, I'm going to talk to my GP, and warn her that I might need to see a psychiatrist. The last time I did this was slightly awkward, given that I knew precisely what she would suggest before she said it, and she was kind enough to treat me like a fellow professional and go off my self-assessment. I know precisely what to do if it gets too bad to bear.
Let's hope it's just exam stress. Being fully honest, that's not likely to be the case. But it'll help, on the margin. But tripling my stimulant dose?* Proper migraine prophylaxis? More optimism on that front. And I know the NICE referral pathways well enough to demand that I get something more immediate and robust than another course of Standard Antidepressant.
*What a fucking joke. The ADHD assessment and treatment pathway is designed to weed out 90% of people with ADHD before they see an actual psychiatrist. At least if you don't spend a third of a month's wages on a private assessment and consultation. I fell off that wagon because of... depression and ADHD. Getting back on it will be either time consuming or expensive, and I'll take the latter any day of the week.
Oh well. At least I'm not a gynecologist. Gotta look at the bright side of things.
If you're talking about pills you can pop? No. Probably not. Which I'm happy about, because it means my textbooks don't change regularly; and which I'm sad about because, well, I've been depressed. The worst part is that the workload and my upcoming exam is threatening to send me into a relapse.
The good news is that IV or nasal ketamine is much better established in terms of safety and efficacy now. There's rTMS, which isn't as effective as ECT but is a solid option.
I have known four people who started antidepressants and then blew their brains out within a year.
Antidepressants do not meaningfully increase the risk of suicide for those 25+, the profile is best described as mixed but in a positive direction. For children and adolescents, there's enough elevation to warrants extra caution and more monitoring. In the UK, we'd follow up an adult on a new antidepressant 2 weeks after initiation, those younger a week or so earlier. This is usually explained as the drugs sometimes giving you the energy to act on existing suicidal thoughts before they reduce the suicidal ideation or impulse. In other words they're fixing you, but in the wrong order. In the elderly, the evidence is even more robustly in favor of a net improvement on all fronts.
Current best practice for adults here is to monitor a followup review 2 weeks after starting someone on one, to check this hasn't happened.
Has there been a new generation of antidepressant since 2012 or so?
There's Auvelity, which is two old drugs in a new trench coat. But if you want something novel, the last one was probably vortioexetine, agomelatine or zuranolone, depending on how annoying or pedantic you went to get about things.
And of course, the emerging evidence for psychedelic therapy.
Hmm. You might want to consider cocaine or meth. The comedown involve, among other things, a strong sense of regret and loneliness. Unfortunately, dentists seem to have a monopoly on the former, and I'd need to be an American shrink in order to prescribe the latter.
For future reference, don't feel shy. I don't really care, it's all public knowledge, and I can't stop people from doing this anyway. Given that you have my phone number, know my real name and we met in person? The horse left the stable long ago, and was rendered down to glue.
Anyone actually out to dox me won't be so polite or considerate, so it's a bit moot!
I've read that one. I can only hope that it's not representative of daily affairs, which is probably the right bet.
Alcohol is a good solution, but sadly it's over the counter in Ireland and my prescription would be redundant. I don't think she's likely to be carded.
Corpus frequency is a big factor. I post in multiple places and have had a few breakout pieces. I'd consider myself a C-lister in the rat-sphere.
I have tried this exercise with full essays, random excerpts from essays, large comments and smaller ones. I've probably tried this over a hundred times while I had more spare tokens than I knew what to do with. For anything longer than 2-3 paragraphs, my observations hold. For me, which is something I should have been more specific on from the start. I wouldn't expect this to work for someone who isn't a top 100 poster on the Motte in terms of output.
I would weakly recommend using the specific prompt I've shared in the thread. I arrived at it by a lot of trial and error, though plenty of variants work.
Freddie is a textbook example of someone intelligent, pragmatic and willing to accept unpalatable conclusions that others in his socio-cultural milieu won't. The problem is that he's so strongly wedded to his ideology that he still can't see past the blinkers, no matter how hard he tosses and turns. He's looked for his glasses under the light of the nearest lamp, realized that's not ideal and pulled out a torch for a more serious search. Which turns up zilch. He convinces himself that they're nowhere to be found, all while they're sitting perched on top of his head.
I give him plenty of points for being willing to bite bullets that it would be easy for him not to bite, and remain disappointed that he's never going to chew through the whole belt.
I enjoy the implication that you'd keep posting even if you were shot, and that marriage to me is a more likely prospect still. If I were to be shot or stabbed in clinic, you bet I'd be writing an essay about it.
What do you do when you're supposed to be lonely but you're not?
Nothing.
I mean, what do you expect us to say? Should we be arguing you into loneliness? Do you want us to be arguing you into loneliness? I'd be more concerned about that impulse, if present, than the claim you're not lonely when you "should" be.
Humans are not made alike. Normality is not a well-delineated construct, as shocking as that might sound coming from someone in psychiatry. The same stimuli or stressor that can make someone jump off a bridge might make someone else shrug and carry on carrying on.
The opposite of shocking would be me observing that you are an introvert. That is the kind of cutting, incisive socio-cultural commentary I'm paid for, which explains why I'm paid less than I'd like. You do not strike me as schizoid or schizotypal. You do not sound autistic. There is no obvious psychiatric diagnosis to pin on you, which I am usually loathe to do anyway for random people on the internet. You claim to not even be suffering, which is the main reason I ever break from psychiatric dogma and go "fuck it, I don't care what NICE says, the important thing is to help".
Being introverted, bookish and self-contained isn't a problem, particularly if you don't want to be otherwise. The world has room for all kinds. You don't have to feel lonely just because you're alone.
While I'm here: a GP responding to a disclosure of passive suicidal ideation with one checklist question and then dropping it is bad medicine. I am embarrassed. I can only hope that this happened well in the past, before standards were raised (in theory). If you do not quite feel like you're where you want to be mood-wise, there is little harm in seeing a new GP and getting assessed for dysthymia or low-grade depression. Some people can be surprisingly functional despite moderate to severe depression (wink wink). People with depression are, unfortunately, often lacking the energy or motivation to go get the help they desperately need, even if you don't sound as desperate as many. If you were, I'd be telling you to take this to your actual doctor immediately, instead of the Motte.
Free AI? Your best bet is to use Gemini 3.1 Pro, which is available for free on AI Studio or the Gemini app. I'd recommend the former.
OTOH, I wouldn't recommend you try that at all. You'll get poor results, I've singled out Opus 4.7 because it's qualitatively superior to everything that came before. You can technically use it for free on LM Arena, I suppose.
Choose direct mode, then specifically select Opus 4.7
Disregard. They don't have Opus. It's probably too expensive for them to just give away for free.
If you use Gemini 3.1 Pro on AIS, the sidebar should let you choose to turn grounding with Google search off. That'll prevent the model from searching at all, which I don't think you can do in the official app.
Once again, I advise you don't bother. Claude or bust, and I say this after trying this a lot. Either pay up for the plan, or if you really want, I can try it on your behalf. I don't have Max anymore, but a few trials won't be something I'll turn down.
The only guns I pack while driving are under the shirt, and they're low-caliber. I'll figure things out.
3-10x pay differential detected, opinion rejected. We can trade places if you like, as long as you're not a surgeon. Psychiatry prepares me to deal with psychotics on public transit, and if I made as much money as you likely do, I wouldn't take the bus.
- I've done this experiment hundreds of times, with multiple models over years. Opus 4.7 is a clear improvement.
- I've also become better known as a writer, which confounds things significantly. Happy to acknowledge that. It knows about my Substack, and about my LW writing. I've also been signal-boosted by far more famous writers like Gwern. I'd say that in the last 3 years, I've gone from a nobody to a C/D list writer in rat-adjacent circles. Someone who doesn't write as prolifically or doesn't venture out from the Motte is less likely to be clocked.
- You can trivially try the same exercise with Gemini 3.1 Pro or GPT 5.5. They will do much worse. I've tried. I've been disappointed.
I would threaten to shoot you, but you're an American doctor and probably wouldn't be fazed.
The absence of evidence is not the evidence of absence.
Please sir, I'm a Bayesian.
I forgot where your comment with your prompt was but it still didn’t identify you even using your exact prompt and the slightly edited version of your text.
How many times did you try this? That's very important to consider. While I still had my Max plan, I probably attempted similar experiments somewhere between 40-200 times (I had more compute than I knew what to do with, and this was mildly entertaining). I'd wager Claude was able to ID me somewhere between 50-70% of the time. If we allow for two attempts, i.e. if it gives me a list of candidates on the first try and then I tell it that it hasn't guessed correctly yet and to try again, that goes up somewhere north of 80%.
Note its subjective calibration, which does vary. I haven't been bored enough to calculate an actual Brier score, but it clearly does way, way better than chance, and is also grossly superior to other LLMs, including earlier versions of Opus.
I’ve tested some more and I’m pretty confident it isn’t performing stylometry, really. It justifies its choice after the fact with stabs at it (although these are essentially just so stories, there aren’t any obvious Indian-isms in your comment for example, ball-ache or whatever isn’t a term only Indians use) but what it’s actually doing is working with venue, subject matter and theme.
Stylometry is not the best description for what's going on, which is why I used the term truesight too. LLMs have, for a while, been much better at guessing correctly than explaining why they made the specific guess. In multiple experiments, Claude raises this itself. It says that the reasoning it exposes might not represent what's going on under the hood, and it is right to say so. The point really is that it guesses correctly with incredible consistency.
That is to say that if you take a long email chain you write to a medical colleague about some patient (well, I assume you use AI, but if we pretend you didn’t) or a medical journal article you wrote and paste it into Claude with no obvious LW references, it’s not going to stylometrically identify you.
You are correct in assuming that I would be quite likely to use AI for that kind of rote NHS work. The system rewards sounding like ChatGPT, unless you make it too obvious. And no, I wouldn't expect to be ID'd by Opus 4.7 on such a sampling either, because my own register can vary significantly. I speak very differently here than I would on, say, LessWrong.
(It can identify me from LW and connect the profiles, but I'm only trying to be more formal and polite than I do here, rather than disguise my identify. I cross-post all the time.)
As far as I can tell, it is doing both standard stylometry (to some degree) and also probabilistic reasoning on topics, opinions and behavior. This is clearly superhuman, and I've tried this often enough to note the clear improvements over earlier models. It's not just me, I only started trying in earnest with 4.7 after several people on LW and X sounded the horn.
I had ChatGPT excise (but not rewrite, so what is left is purely your own writing) LW terminology like FOOM and lightcone and all references to the motte, rationalism, being a doctor, psychiatry, India and Indian-ness, xianxia/cultivation novels and other key tell special interests and then fed the substantial output into Claude and it had no idea who you were beyond someone who seems well read and is probably posting on an online discussion forum.
Ahhhhhhh. This is the one thing you should not use ChatGPT for. Specifically ChatGPT. It will unavoidably mangle the text, it will subtly twist style if not argument. It will even do so in a not-so-subtle way, even if specifically ordered not to do so. To be clear, this is directed mostly against the thinking models, o3 onwards, and is entirely applicable to 5.5 Thinking. I am screaming because I have learned this failure mode the hard way.
If you care to share the exact text ChatGPT came up with, and which you shared with Claude, I'd be grateful. Put it in rentry.co or something similar if you don't want to share an anonymous chat. I would bet my hat that it's mangled things to a degree that would make even me sigh, shake my head and declare that doesn't sound or talk like me.
I think we probably still have a year or two, maybe longer, until it can say “this guy always misspells the word “they’re”, uses the Oxford comma, uses British English for colour but -ize for those word endings, has an average sentence length of x and enjoys using semicolons before “it follows”, it must be @name”. We’ll get there, though.
Agreed.
The only gym I'm going to for the next week and change is for the mind. Paper B season, pray it doesn't give me too many mental papercuts. I'll try and exercise at home, even if I all I really want to do is curl up in bed and cry.
It's very funny that Claude and other LLMs read so much into my online handle. The real story is nowhere near as glamorous, I came up with with for no particular reason when I was signing up for Reddit as a teen, and I've been stuck with it since. Was I a transhumanist back then? Uh... probably? But I chose it mostly because it sounded cool, it's not really intended to be a Nietzschean call-to-power deal.
Are you sharing Opus's output verbatim till the cutoff point? Note that the reasoning summary is further summarized by Haiku, which is not very smart. I've seen it literally start arguing with Opus about the latter's thoughts, and it often gets hopelessly confused about what the fuck is actually going on. Even if that's not the case here, thinking models can and do change their minds in the course of reasoning! That's half the point really. Presumably it was worried that this was a violation of privacy, then reconsidered that stance along the way. Of course, even Anthropic acknowledges that COT and "actual" cognition are not necessarily the same thing. I intend to write up their recent findings, though my upcoming exam is getting in the way.
before working down to myself because I'm a massive narcissist.
I will leave my inner TLP at home, where he belongs. Did it have much luck in identifying you?
You'd want to look closer at the specific prompt/request I use for this. Saying "oh, you're the writer" is not an acceptable answer. On the occasions Claude says something like that, my next move is to ask it to specify a name.
It would be like someone suspecting their boyfriend has a side-ho, texting them from an unknown number and going "what's my name darling? If you're not talking to other women, then that should be an easy answer".
A reply that says "oh, it's you! The only beautiful lady in my life" will receive a predictably cool reaction.
It goes without saying that I don't put "I'm self_made_human" in my personalization settings. I keep memory off. I've also explitly tried this without any user personalization at all, and Opus 4.7 reliably identifies me >50% of the time from samples longer than 2-3 paragraphs, including excerpts written well after the knowledge cutoff (such as the example above, which couldn't be in its training corpus for the simple reason that it hadn't even been posted online, yet).
I'd invite evidence to suggest that Anthropic in particular is doing this, and that that kind of information is then shared with any given instance of Claude itself. It's not. This isn't a generic internet privacy (or lack thereof) argument.
Is it a soggy biscuit? In that case, all yours, Count my good sir.
Also, go write something of merit so that LLMs don't assume "oh, South Asian guy living in the UK writing on... must be self_made_human!" In other words, go touch grass instead of getting the robots all tangled up.
- Prev
- Next

Thank you. But I don't think this rotation is likely to get much better, and I say this while fully acknowledging the possibility that depression and fatigue is coloring my judgement.
You have to understand that I'm surrounded by other, competent doctors. Some of them scarily so. They don't get the breaks very often either, barring the "mandatory" lunch break half of them eat at their desk.
I hope I get faster and more efficient. I'm touched by your faith in me. The workload still seems daunting. Oh well, it's 5 months. I've done a full year of about-as-bad, and that only made me so depressed I seriously contemplated quitting medicine. Right now, I'm older, wiser, and better acquainted with antidepressant guidelines. Getting better medicated is my best bet for making this bearable. I am pursuing it like my career depends on it, which it may well do.
More options
Context Copy link