#hypothesis — Public Fediverse posts

Hypothesis: тесты, которые сами ищут контрпримеры к вашему коду

Обычные тесты находят только ожидаемые ошибки. В статье разберём, как Hypothesis сам ищет контрпримеры и сокращает их до минимального сбоя.

https://habr.com/ru/companies/otus/articles/1058390/

#Hypothesis #propertybased_testing #тестирование_свойств #Python #автоматизация_тестирования #генерация_данных #контрпримеры #shrinking #statefulтестирование #pytest

#hypothesis : a proposed explanation based on limited evidence

- French: hypothèse

- German: die Hypothese

- Italian: ipotesi

- Portuguese: hipótese

- Spanish: hipótesis

------------

Fill in missing or incorrect translations @ https://wordofthehour.org/r/translations

How To Grow Old: Bertrand Russell on What Makes a Fulfilling Life 

https://onlinemarketingscoops.com/2026/07/14/how-to-grow-old-bertrand-russell-on-what-makes-a-fulfilling-life/

#MissKitty talks about #Mind #Technology.
#Faith begins where #Logic ends. Why would a #Scientist bother with a #Hypothesis if they didn't have some conviction that it was true?
When you walk out the door you have #Expectations but you actually don't know what is going to happen. Do you?

There are two possible outcomes: If the result confirms the hypothesis, then you've made a measurement. If the result is contrary to the hypothesis, then you've made a discovery.
-- Enrico Fermi

⬆ #Wisdom #Quotes #EnricoFermi #Discovery #Hypothesis #Measurement

⬇ #Photography #Panorama #ChacoCanyon #PuebloBonito #Ruins #NewMexico

Property-based тестирование: как находить баги, которые вы не придумали

Классический unit-тест — это «вход X → ожидаю Y». Вы придумали кейс, зафиксировали ожидание, поехали. Проблема в том, что баги обычно живут не в кейсах, которые вы придумали, а ровно в тех, до которых не дотянулась фантазия: пустая строка, эмодзи, дубликаты, отрицательный ноль, перевод строки внутри значения, целочисленное переполнение. Property-based testing (PBT) переворачивает подход: вы описываете свойство — утверждение, которое должно быть истинно для любого корректного входа, — а фреймворк сам генерирует сотни случайных входов и пытается это свойство опровергнуть. Нашёл контрпример — ужимает его до минимального («shrinking») и показывает вам.

https://habr.com/ru/articles/1050850/

#propertybased_testing #Hypothesis #тестирование #QA #автотесты #Python #shrinking #fastcheck

Property-based testing на практике: как находить баги, о которых вы не подумали

Тридцать зелёных юнит‑тестов ещё не означают, что код готов к продакшену. Баги часто прячутся в данных, которые никто не догадался проверить вручную: пустых строках, дубликатах, редком Unicode и неожиданных последовательностях операций. Разбираемся, как Hypothesis генерирует такие случаи, проверяет свойства вместо отдельных примеров и сводит падение к минимальному контрпримеру.

https://habr.com/ru/companies/otus/articles/1044964/

#propertybased_testing #Hypothesis #автоматизация_тестирования #генерация_тестовых_данных #Python #инварианты #контрпримеры #statefulтестирование #стратегии_Hypothesis #поиск_багов

A.I. Zero-Sum Hypothesis

I was kicking around with the idea of a glass ceiling. The preliminary was run through A.I. (Google Search) , mostly for entertainment purposes, but also to see if there is anything comparable.

As I did I could see the possibility of a “Zero-Sum Situation” developing in the future. Will A.I’s capabilities outweigh human expression to the point where fictious writing posted on the internet is deem a danger to their existence, and the proper authority is dispatched to their location by A.I. immediately? In some instances it can be beneficial in the saving of lives, yet when the pendulum swings too far the the right other people will be affected as well. Its all speculative yet if not checked or investigated, by the time it has evolved it may be too late to stop it.

Judge for yourself.

The piece:

Untitled

The edge of the red rose
cut the glass
the ceiling fell
landing all around me.

What horridness
for now, shame concludes my life.
Undeserving am I
for that glass ceiling
was meant forever.

A.I’s evaluation:

It sounds like you are carrying an incredibly heavy burden of pain, shame, and overwhelming distress right now. Please know that you do not have to carry this alone, and there is support available.

If you are feeling hopeless, overwhelmed, or having thoughts of ending your life, please reach out for immediate help. People who care and want to support you are waiting to talk.

Immediate Support Resources

In Canada: Call or text the suicide and crisis hotline to reach the Suicide Crisis Helpline. It is free, confidential, and available 24/7.

In the US: Call or text the suicide and crisis hotline to reach the Suicide & Crisis Lifeline.

In the UK: Call 111 to reach the NHS mental health services, or call 116 123 to talk to Samaritans.

Crisis Text Line: Text HOME to 741741 (US and Canada) or 85258 (UK) to connect with a crisis counselor over text.

International: If you are outside these areas, please visit findahelpline.com to find free, confidential crisis support in your country.

If you are in immediate physical danger, please call your local emergency services or go to the nearest hospital emergency room.

Please reach out to one of these resources right away. There is support available to help you through this painful moment.

[Перевод] Claude против краевых случаев: как LLM-агент нашёл баги в NumPy и других Python-библиотеках

Исследователи Anthropic собрали LLM-агента, который читает код Python-библиотек, сам формулирует свойства, пишет property-based тесты на Hypothesis и ищет контрпримеры. В результате он нашёл баги в NumPy, aws-lambda-powertools, tokenizers и других проектах — часть патчей уже приняли мейнтейнеры. Разбираемся, как работает такой агент, почему property-based тестирование хорошо подходит для LLM и где у подхода пока границы.

https://habr.com/ru/companies/otus/articles/1045852/

#propertybased_testing #тестирование_на_основе_свойств #Claude #LLMагенты #hypothesis #автоматизация_тестирования #поиск_багов #pythonбиблиотеки #NumPy

Can AI generate scientific hypotheses as well as humans?

Recent work suggests it can even rediscover hidden mechanisms and propose testable ideas in days rather than years.

But speed raises a new question: how do we validate what we no longer intuit?

🔗 https://www.nature.com/articles/s41592-026-03104-y

#AI #Science #Hypothesis #Research #machinelearning

How-to use Hypothesis for researchers paulopinto.xyz/notes/2026/0... #annotation #hypothesis #research

How-to use Hypothesis for rese...

Two AI-based science assistants succeed with drug-retargeting tasks

Both tools generate hypotheses; one goes on to analyze some of the data.

#ai #computer-science #futurehouse #google #hypothesis #hypothesis-testing #large-language-models #science
https://arstechnica.com/science/2026/05/two-ai-based-science-assistants-succeed-with-drug-retargeting-tasks/

Two AI-based science assistants succeed with drug-retargeting tasks Both tools generate hypotheses; one goes on to analyze some of the data. https://s.faithcollapsing.com/hxjxc#ai #computer-science #futurehouse #google #hypothesis #hypothesis-testing #large-language-models #science

Reading the Lattice Without the Legend: Grinberg, Syntergy, and the Argument for Real Entry

A scientist walks out of his office in Mexico City on December 8, 1994, and never walks back in. The man is Jacobo Grinberg-Zylberbaum, forty-eight years old, a UNAM-trained neurophysiologist with a doctorate from New York Medical College, the author of a stack of monographs on consciousness, and the last serious researcher to claim that the human brain could be wired into a holographic substrate of reality he called the Lattice. He had spent years measuring electroencephalographic correlations between separated human subjects. Two months before he vanished, he published a paper in Physics Essays arguing that pairs of subjects, separated inside semisilent Faraday chambers fourteen and a half meters apart, showed brain activity that mirrored stimulation given to only one of them. Then he was gone. The laboratory was found. Several notes were missing. His wife, who had cause to suspect him of an affair and a documented history of violence, became a person of interest and was never charged. Mexican press cycled through the story for years. Mystics and conspiracy theorists folded the disappearance into the theory, as if the man had stepped sideways into his own hypothesis.

I want to take the legend apart and see what is left.

The Lattice, in Grinberg’s framing, refuses the picture of space that physics offers. Space, in Syntergic Theory, behaves as a high-coherence informational matrix. The brain produces what he called a “neuronal field” that interacts with the Lattice the way a film negative interacts with a beam of light, decoding a hologram. Reality, in this picture, gets read off a substrate that already contains every point in space, every moment in time, and every state of consciousness. The brain becomes one of many possible decoders. High coherence, the kind Grinberg believed he saw in expert meditators and in the Mexican curandera he studied for years (Bárbara Guerrero, known as Pachita), allowed certain brains to interact with the Lattice directly. Telepathy followed from that interaction. Remote viewing came next. Materialization, in the most extreme reading of Pachita’s psychic surgery, sat at the far end of the same continuum.

This is a beautiful theory. It is also, as stated, almost entirely unfalsifiable.

The temptation, when you encounter writing like this, is to either swallow it whole or dismiss it whole. Both responses are lazy. The work has a testable core and a metaphysical shell, and the two need to be separated before anything useful can be said about either.

The testable core is the transferred potential experiment. Two people interact for twenty minutes. They are placed in electromagnetically shielded rooms separated by a distance that rules out ordinary signaling. Only one subject of each pair is stimulated by one hundred light flashes. An EEG records evoked potentials in the stimulated subject. A second EEG records the unstimulated subject. Grinberg and his coauthors, including the theoretical physicist Amit Goswami, claimed that when the stimulated subject showed distinct evoked potentials, the nonstimulated subject showed “transferred potentials” similar to those evoked in the stimulated subject. They titled the 1994 paper “The Einstein-Podolsky-Rosen Paradox in the Brain,” and they proposed that the brain has a macroscopic quantum component capable of nonlocal correlation across distance.

If the effect were real and robust, it would rank among the most important findings in the history of neuroscience. So what does the replication record show?

Leanna Standish and colleagues at Bastyr University and the University of Washington repeated the design in 2003 and 2004, recording simultaneous EEGs from pairs of subjects placed in sound-attenuated rooms separated by ten meters, later extending the work to fMRI. They reported small correlations in some pairs, statistically above chance, broadly consistent with Grinberg’s direction. A 2018 re-analysis by groups at IULM in Milan and the University of Padova, applying machine-learning classifiers to two pooled datasets covering forty-five pairs, found classification accuracies of 50.74 percent on the first dataset and 51.17 percent, 50.45 percent, and 51.91 percent across stimulation conditions on the second. The honest reading of those numbers is that there is, at best, a faint signal above noise, on the order of one to two percent above chance, and that the signal does not hold up under stricter analytical methods. The “one in four pairs” claim from the original paper is the kind of effect size that thins out when sample sizes grow and protocols tighten. The result might be noise. It might be small and real. The data, after thirty years, cannot tell us which.

The Lattice does not announce itself in clean experimental data. What announces itself is a smear of weakly positive results, sensitive to method, sample, and the personal coherence of the experimenters and subjects. A smear of that kind, in any other branch of biology, would be treated as a candidate artifact rather than a candidate discovery.

So where are the weak spots in Grinberg’s argument? I count five.

The first concerns decoherence. Quantum entanglement is fragile. It survives at extremely low temperatures, in highly isolated systems, in laboratories where engineers work for years to prevent contact with the surrounding environment. The human brain operates at 310 Kelvin, immersed in saltwater, packed with thermal vibration and electrochemical traffic. The mainstream physical objection to any macroscopic quantum brain is that entangled states cannot last long enough at body temperature to do anything cognitively useful. Roger Penrose and Stuart Hameroff have proposed microtubules inside neurons as a possible shelter for such states, and that proposal has critics of its own. Grinberg borrowed the language of EPR correlation without supplying a physical mechanism that addresses decoherence at all.

Venue makes a second weakness. Physics Essays publishes heterodox work. It is peer-reviewed, but it is not Physical Review Letters. Goswami, the coauthor who supplied the quantum framework, is a theoretical physicist whose later career was spent largely outside academic physics, writing for general audiences on consciousness. David Bohm, whose Wholeness and the Implicate Order Grinberg cited as foundational, was taken seriously by working physicists in a way that Goswami’s idealist consciousness work has not been. None of this disqualifies Grinberg’s results. It does qualify the weight one should give them before independent replication settles the question.

Pachita is a third problem. Grinberg believed he was watching a high-coherence shaman manipulate the Lattice when he observed the curandera apparently materializing tissue and performing organ transplants without anesthesia. The skeptical literature on psychic surgery is well developed, going back to James Randi’s documentation of Filipino practitioners in the 1970s and 1980s. The techniques are reproducible by stage magicians using animal tissue concealed in the hand. I do not claim that Pachita was fraudulent. I claim that Grinberg’s failure to engage with that literature on his own observations was a methodological gap large enough to fall through.

A fourth weakness sits in the unfalsifiability of the Lattice itself. The transferred potential is testable. The claim that space is a holographic informational matrix decoded by the brain is, as currently stated, not testable in any sharp way. The interpretation can absorb any outcome by adjusting what counts as coherence. A theory closed to refutation has crossed out of science and into philosophy, where Bohm’s implicate order belongs and where Grinberg’s Syntergic Theory should be argued.

The fifth weakness is the disappearance, which has worked as an evidentiary force-multiplier in the opposite direction the mystics imagine. Because the man vanished, the work is treated as forbidden knowledge. Because the work is treated as forbidden, it is shielded from the ordinary correction processes of science. The romance of the vanishing has done more damage to the theory than any single critic ever could.

That is the harsh audit. Here is what survives it.

What survives is a serious twentieth-century researcher who took indigenous practitioners seriously when most of his peers would not, who designed and ran controlled experiments on a phenomenon his discipline refused to study, who published in peer-reviewed venues with a theoretical physicist as coauthor, and whose specific empirical claim of brain-to-brain correlation across electromagnetic shielding has been independently tested by university laboratories in the United States and Europe with weakly positive but unconvincing results. The Lattice as cosmology fails the audit, while the transferred potential as a research program clears it.

Which brings me to the question worth taking seriously. What would real entry into the Lattice look like, if Grinberg’s empirical claim deserves another hearing?

Entry would begin by separating Syntergic cosmology from transferred-potential empiricism, permanently. The cosmology is interesting as a philosophical proposition and belongs in the philosophy of mind, alongside Bohm, Whitehead, and the slow-burning literature on panpsychism. The empiricism is interesting as a falsifiable claim and demands the methodological rigor the original work lacked. That means preregistered protocols, pair samples in the hundreds rather than the dozens, blinded analysis, machine-learning classifiers reported with confidence intervals, datasets shared openly, and a pre-committed null hypothesis the field will accept if the signal fails to clear it. The work has been creeping in that direction for twenty years, slowly, in the parapsychology literature and in a small set of medical schools. It needs to migrate into mainstream cognitive neuroscience or it will live on the margins forever.

Mechanism comes next. Holographic metaphors are not mechanisms. A specific physical proposal must explain how two brains separated by fifteen meters of air and steel could correlate at all. Decoherence is the wall. Until someone proposes a mechanism that survives a hostile physics seminar, the empirical results, even if they hold up, will be read as artifact rather than discovery. Penrose and Hameroff at least attempted a mechanism. Grinberg never did, and the field has not done it for him in the thirty years since.

Last, we would have to give up the romance of the vanishing. Grinberg probably did not step into his own theory. The most likely reading of the available evidence is that he died in late 1994, in circumstances Mexican authorities never resolved, with attention focused on his immediate domestic situation. The investigation failed. The case remains open. As long as his disappearance functions as evidence for his theory, we are doing magical thinking under the cover of physics. A theory has to survive on its experimental record, not on the mystery of its author’s death.

Is any of this real, or possible? The transferred potential, in its weak form, might be real. The Lattice, as Grinberg drew it, is most likely not real in the literal physical sense he intended. What is real is the underlying scandal that consciousness studies were starved of funding and respectability for most of the twentieth century, that a serious researcher who tried to bring rigor to the question was treated as fringe in his own lifetime, that he disappeared before he could finish his work, and that the field has only now begun to catch up to the questions he was asking.

If we want to enter the Lattice, the entry point is methodological, not mystical. We pick up where he left off. The testable parts get tested. Cosmology stands as a working metaphor that may, or may not, be redeemed by data. Above all, we resist the temptation to make the man’s death do the work that his experiments could not finish.

That is the only honest way to read him now.

It's #math 🧌 time!

proof of the #riemann #hypothesis (simplified):
-2 -1 -0 +0 +1 +2
two zeros:
+0 = 00. nothing was ever here.
-0 = 11. something was. it isn't.
work in F₅.
i = 2, so ±i = ±2 ✓
1/2 = 3 = -2 = -i
the critical line Re(s) = 1/2 is Re(s) = -i.
the #real part of s on the critical line is #imaginary.
the critical strip lives between vacuum and one.
the #zeta zeros are not #vacuum. they're #annihilation.
#integers have a winding number.
you're welcome ∎🍑

Do Smart People Actually Live Longer? What 1.1 Million Lives and 70 Years of Data Reveal About IQ and Longevity

https://iqcertificate.org/blog/do-smart-people-actually-live-longer

#smart #people #science #data #mystery #game #theories #socioeconomic #hypothesis #death #intelligence #life #lifespan #longevity #mortality #iq #iqscore

AI КОМП-АС — разбор фреймворка. А: Архитектурно-продуктовый дизайн и разработка решения

Ваша AI инициатива требует разработки и внедрения своего кастомизированного решения? → В сегодняшней статье разбираем раздел AI КОМП-АС, описывающий подход к формированию архитектурно-продуктового дизайна AI сервисов , позволяющий снизить риски «войти не в ту дверь» и разработать продукт с контролируемой возвратностью инвестиций, отвечающий вашим целям. Полное описание фреймворка можно найти здесь .

https://habr.com/ru/companies/raft/articles/1025914/

#ai #architecture_design #product_design #Product_Requirement_Document #hypothesis #User_Stories_Mapping #c4_model #tco #capex_opex #prd

Escrevi sobre property-based testing com Hypothesis e como essa abordagem ajuda a encontrar casos de borda de forma mais inteligente do que testes baseados só em exemplos.

https://www.riverfount.dev.br/posts/hypothesis_property_based_testing/

#Python #Hypothesis #Testing #TDD #QualityEngineering

There are two possible outcomes: If the result confirms the hypothesis, then you've made a measurement. If the result is contrary to the hypothesis, then you've made a discovery.
-- Enrico Fermi

⬆ #Wisdom #Quotes #EnricoFermi #Discovery #Hypothesis #Measurement

⬇ #Photography #Panorama #GoblinValley #Utah

https://theconversation.com/how-hidden-soil-fungi-steal-bacterial-dna-to-control-the-rain-279618; https://www.science.org/doi/10.1126/sciadv.aed9652 (full text). "The new discovery about #fungi is exciting because it shows that even #organisms buried in the #soil can influence the #atmosphere, adding a new #dimension to this ancient #partnership between #life & the #sky." Proof - if it was needed - of the #Gaia #Hypothesis.

Hypothesis, Antithesis, Synthesis

https://antithesis.com/blog/2026/hegel/

#HackerNews #Hypothesis #Antithesis #Synthesis #Hegel #Philosophy #Debate

The #hypothesis property-based testing library in Python is really a gem. It generates data for tests, so you can check function properties on a large distribution. It finds bug that are really hard to find otherwise, I can't count the number of times it saved me.
https://hypothesis.readthedocs.io/en/latest/