Looking back at my own education, I realize I never really understood what grades meant.
In school I always had straight A’s in math - through my 'math/physics magnet middle/high school' and the early university years. Then, near the end of my undergraduate studies and into graduate level, the abstraction level finally outran me. What once felt easy became a fog of symbols and proofs I could barely follow. My grades dropped from A’s to barely passing.

Now my son is in middle school, facing math tests that seem simple compared with what I used to do, and yet he usualy brings home B’s and C’s. At first I was worried he was struggling. But after thinking about how math actually builds from one layer to the next, I began to see grades differently.
Math is cumulative: arithmetic => fractions => basic algebra => trigonometry => calculus => abstract algebra and other abstract branches.
To pass any stage, even with a C, a student must already handle the lower levels - fractions, negative numbers, operations, with near-A competence. In that sense, every “passing” grade is built on a pyramid of earlier mastery. A C in algebra implies solid command of arithmetic; otherwise the algebra would collapse.
And mathematics is abstract all the way down. Even counting natural numbers is an act of abstraction - something crows and primates can do. As we climb the ladder, the required abstraction doesn’t suddenly appear; it grows, from trivial symbolic mapping to ever more demanding mental models. And eventually each person reaches their ceiling - the point where abstraction exceeds what their working memory and reasoning capacity can handle.

This leads to what I see as a two-axis model of learning math. One axis is vertical, measuring mastery of the current/latest subject - the top of the stack of skills built layer by layer. The other is horizontal, measuring the mind’s abstraction capacity - how far one can extend that structure before it becomes too complex to hold. Grades only measure the first axis. The second one, which defines how far a person can ultimately go, remains totally invisible.
Piaget hinted at this when describing the “formal operational” stage — the onset of abstract reasoning — but reality is more continuous. The ability to handle higher abstraction isn’t a stage that suddenly starts; it’s a spectrum that stretches through life, unevenly across individuals.

And here lies the problem: all education systems remain centred on test scores that capture only the vertical axis. It rewards fluency and accuracy but ignores the horizontal reach of abstraction that truly defines mathematical potential. Until education finds a way to see and nurture that dimension, grades will continue to tell only half the story, and I would argue not the most important one.

I am still spreading antivax agenda online, and here is a very common counter-argument I am often confronted with. An official USA CDC publication: Cardiac Complications After SARS-CoV-2 Infection and mRNA COVID-19 Vaccination presumably proving that heart inflammation from covid is more common than one from the vaccine in both sexes and each age bracket.

When reading this paper first I thought that they compare heart inflammation risk(HIR) in the vaccinated groups vs unvaccinated groups, and proved that the latter is bigger. That would certainly change my mind about the subject! Unlike covid vaccination enthusiasts who just believe in science, my beliefs can be changed with new evidence.

Then I realised they don't compare HIR/vac vs HIR/unvac groups, I thought that maybe they compare HIR/vac vs HIR/(unvac&infected). No, it is not quite the case either.

What they compare is (HIR)/vac vs (HIR+MIS)/(positive covid PCR test), and latter is larger. Which makes total sense because not every covid infection gets PCR tested in hospital or testing site, especially with Omicron+ versions of the virus. So the second fraction gets significantly larger because of small denominator.

So they are correct: the HIR from vaccination is less common than HIR from severe covid. But parents are more interested in HIR from vaccination vs HIR from not vaccinating, which is entirely different Bayesian reasoning game.

How can I estimate/quantify the difference? The numbers in the article are the following: "Among males aged 12–17 years, the incidences of myocarditis and myocarditis or pericarditis were 50.1–64.9 cases per 100,000 after infection, 2.2–3.3 after the first vaccine dose, and 22.0–35.9 after the second dose; incidences of myocarditis,
pericarditis, or MIS were 150.5–180.0 after infection. So, TEEN BOYS (ages 12-17 years) had 2-6X the risk of heart complications after infection compared to after vaccination."

However, noting severity of Omicron+ in children, more than 80% of the cases were asymptomatic or so mild that no parent would bring a child for PCR testing, at most there is just at-home antigen test used. (According to MMWR seropositivity study by February 2022 ~55M children were seropositive and 12M children tested positive in US) This makes second fraction 5x smaller if we count all infections.

Which makes final comparison show that vaccination is 0.8x-2x more dangerous than covid infection for young boys. (Assuming the chance of infection is 100%, which is also not the case).

Maybe that is why an official FDA representative recently wrote the following words in an opinion piece published in JAMA : "During the COVID-19 pandemic, broad, one-size-fits-all mandates were aggressively pursued, requiring vaccines in populations with potentially net-negative benefit-risk profiles. This ... in part, [may] be responsible for widespread loss of trust in public health and medicine and growing hesitancy against all vaccination. Public trust takes decades to build, but can be forfeited in a single action."

В нашем дачном массиве в 80-х царили инженеры. Народ полностью городской, с деревенскими терминами не знакомый.

Сосед наш, инженер-конструктор Андрей Ильич, решил заняться птицеводством по-научному. Соорудил автоматические кормушки, поилки, птичник утеплил, почти куриная кремниевая долина получилась. Летом приезжал раз в неделю, а цыплята росли сами, будто в лаборатории под контролем алгоритмов.
Осенью Андрей Ильич утеплил птичник и пустил в дело топор: зарезал петушков и большую часть кур, оставил одного на продажу. Слышал он, что за живого петуха на рынке дают дороже. И вот идёт он в воскресенье в райцентр, петуха под мышкой тащит, по пути грибы-моховики собирает (местные, конечно, от них нос воротили - Ай-ай-ай, возле дороги!»).
На рынке у него товар скромный, всего один петух. Есть еще небольшая проблема - Андрей Ильич совсем не говорил по-чувашски, а многие покупатели не очень хорошо говорили по-русски. Но общий язык найти можно. Народ подходит, птицу одобрительно щупает и задаёт один и тот же вопрос:
— Куртопчит?
Андрей Ильич не понял, что за зверь такой "куртопчит", подумал - покупатель цену сбивает, ругает петуха по-чувашски. Поэтому ответил:
— Нет, не куртопчит.
Покупатель развёл руками:
— Так на кой он нужен? И ушёл.
Следующему, по закону итераций, Андрей Ильич выдал противоположный ответ:
— Конечно, куртопчит!
Ему попался весёлый, уже слегка поддатый покупатель:
— А как часто?
— Раз в неделю, честно прикинул инженер, будто говорил про планёрку в КБ.
Покупатель сморщился и исчез. А больше никто не подошел, слух уже пошёл по рынку: «Странный вырас (русский - чув.) продаёт петуха, который кур не топчет. Раз в неделю - и то еле-еле».
Так научный эксперимент по птицеводству закончился маркетинговым провалом.

Jacob's last U11 tournament with is over, and now it is time to buy a longer weapon and start attaching wire to the helmet. This time he only got 6th place; worse than before. And once again, most of dangerous opponents only practiced single move.
The most strange one was a guy who was only using one move: take the priority, and ram the opponent with both blade and helmet, while covering his chest with left hand! With serious judges, this would grant him black cards each time, but now judges only warned him, and I overheard when they were discussing "I should disqualify him but I do not want to spoil fun." Jacob still managed to win.
Another guy was an opponent who weights ~2x more than Jacob and more than any other competitor, and he also practiced a single move: really very strong parry which looked like using a club rather than a foil, and then quick attack while the opponent is distracted. Jacob managed to win 5:0 while feinting against parry and then attacking, worked 5 times in a row.
And then Jacob lost in elimination to a left-handed opponent who rushed to start close quarter combat (while doing so, he scored 3 points and my son scored 4), and then in close quarter fight he scored 6 points and Jacob scored 0. He was begging his coaches to teach him close quarters for a long time, but they always responded that it is too early, and he should not let the opponent that close anyway.
Once the first "ramming" guy learns close quarters fight techniques, he will be very difficult to face.

Here is how buggy messenger delivery status can lead to miscommunication.

Message status: [Sent, Delivered, Read]

My view, after mobile internet went down on my phone at some point:
1. Me: Going home now. [SDR]
2. Julia: We'll be at a grocery store soon, should I pick you up at station A?
3. Me: No, thanks, I'll exit at station B and walk home. [SD]
4. Me: I'll be there in 30 minutes. [S]
5. Me: I'll text you if something happens with the train. []

Julia's view:
1. Alex: Going home now.
2. Me: We'll be at a grocery store soon, should I pick you up at station A? [SDR]
3. Alex: I'll be there in 30 minutes.
4. Me: I am waiting, where are you? [S]
5. Me: FindMy shows you at station B, why? [S]
6, 7, 8, 9

I was considering sending an SMS, but as my 3. was shown as delivered, I did not expect Julia to receive my message 4. only, but not 3. I should probably switch to WhatsApp.

The most prominent LJ biologist recently wrote a long article about Trump's CDC/FDA appointees.

One of his points was "А ещё мракобеса посадили в кресло директора FDA по оценке и изучению биопрепаратов, то есть вакцин и генных терапий. Это хирург Марти Мэкари, известный своей резонансной статьёй о том, что повторная вакцинация от ковида приносит мальчикам-подросткам больше вреда, чем пользы. Правда, среди авторов статьи не было ни одного специалиста по вакцинам или инфекционным заболеваниям."

I think this is the article.

Scinquisitor's point was that all authors of that article were not qualified, hence the article is wrong (though BMJ journal of medical ethics editors still published it)

Makary is 8th author (indeed signed as Department of Surgery, Johns Hopkins University), but he is not the only author from Johns Hopkins University. Another one is from Epidemiology, Johns Hopkins University; another one is from Department of Global Health and Social Medicine, Harvard Medical School; and another from Epidemiology and Biostatistics, University of California San Francisco.

Interestingly, Paul Offit is favourably mentioned in both BMJ article and LJ post :)

Looking through the article, I can't really get it why Makary is "мракобес". Anybody knows why?

I just used local Deepseek-r1-llama70b-q8 model and asked it to translate my short story from a previous post to Russian. Here is the result:
( Read more... )

On December 31st we've moved to a new condo. Here is the view from a window:

We were thinking first that we are choosing between an apartment in Munich close suburb or a house in a village, but finally had to settle for an apartment in a village not too far from Munich in the direction of Alps.

Maybe I am a bit too late, but I just discovered this guy

Not a great literature, but OK.
1. "Кровь на воздух" - LLMs in space opera setting. Darkness - 5/10
2. "Седьмая мапа" Cyberpunk world - high tech, low life. Interesting setting - a cyberpunk city. Darkness - 7/10
3. "Хранители Мультиверсума ", Darkness - 8/10
2 & 3 are part of universe of великого кристалламультивесума.
The author knows how to use genai to do illustrations..

Какие времена, такой и Крапивин.

After Black Forest released the Flux model, I think everyone started trying it and comparing with DALL-E3, Midjourney and Stable Diffusion 3. I am cancelling my Midjourney subscription. (Not that I used it a lot, local SD-3 was mostly good enough for me)

On Mac, Flux just works with the following conda env setup:
Python 3.11.3,
pip install torch==2.3.1 torchaudio==2.3.1 torchvision==0.18.1 einops==0.6.1 numba==0.57.0 numpy==1.24.3 transformers==4.33.1 tqdm==4.65.0 tts==0.21.1
Please use the text encoder and Flux-1-dev.
ComfyUI config and pic:
( Read more... )