Axis of Ordinary
Открыть в Telegram
Memetic and cognitive hazards. Substack: https://axisofordinary.substack.com/
Больше3 810
Подписчики
+224 часа
+87 дней
-330 дней
Загрузка данных...
Похожие каналы
Облако тегов
Входящие и исходящие упоминания
---
---
---
---
---
---
Привлечение подписчиков
сентябрь '26
сентябрь '26
+65
в 14 каналах
август '26
+158
в 18 каналах
Get PRO
июль '26
+140
в 9 каналах
Get PRO
июнь '26
+74
в 14 каналах
Get PRO
май '26
+171
в 24 каналах
Get PRO
апрель '26
+114
в 20 каналах
Get PRO
март '26
+63
в 10 каналах
Get PRO
февраль '26
+67
в 9 каналах
Get PRO
январь '26
+56
в 7 каналах
Get PRO
декабрь '25
+44
в 3 каналах
Get PRO
ноябрь '25
+39
в 8 каналах
Get PRO
октябрь '25
+61
в 20 каналах
Get PRO
сентябрь '25
+46
в 9 каналах
Get PRO
август '25
+33
в 9 каналах
Get PRO
июль '25
+66
в 16 каналах
Get PRO
июнь '25
+49
в 8 каналах
Get PRO
май '25
+64
в 9 каналах
Get PRO
апрель '25
+65
в 14 каналах
Get PRO
март '25
+109
в 6 каналах
Get PRO
февраль '25
+80
в 12 каналах
Get PRO
январь '25
+70
в 8 каналах
Get PRO
декабрь '24
+63
в 11 каналах
Get PRO
ноябрь '24
+80
в 10 каналах
Get PRO
октябрь '24
+94
в 14 каналах
Get PRO
сентябрь '24
+107
в 16 каналах
Get PRO
август '24
+99
в 18 каналах
Get PRO
июль '24
+227
в 14 каналах
Get PRO
июнь '24
+85
в 16 каналах
Get PRO
май '24
+119
в 14 каналах
Get PRO
апрель '24
+110
в 25 каналах
Get PRO
март '24
+193
в 41 каналах
Get PRO
февраль '24
+204
в 30 каналах
Get PRO
январь '24
+800
в 42 каналах
Get PRO
декабрь '23
+216
в 27 каналах
Get PRO
ноябрь '23
+232
в 60 каналах
Get PRO
октябрь '23
+154
в 13 каналах
Get PRO
сентябрь '23
+156
в 0 каналах
Get PRO
август '23
+209
в 0 каналах
Get PRO
июль '23
+302
в 0 каналах
Get PRO
июнь '23
+95
в 0 каналах
Get PRO
май '23
+243
в 0 каналах
Get PRO
апрель '23
+127
в 0 каналах
Get PRO
март '23
+236
в 0 каналах
Get PRO
февраль '23
+36
в 0 каналах
Get PRO
январь '23
+40
в 0 каналах
Get PRO
декабрь '22
+30
в 0 каналах
Get PRO
ноябрь '22
+36
в 0 каналах
Get PRO
октябрь '22
+40
в 0 каналах
Get PRO
сентябрь '22
+23
в 0 каналах
Get PRO
август '22
+35
в 0 каналах
Get PRO
июль '22
+36
в 0 каналах
Get PRO
июнь '22
+43
в 0 каналах
Get PRO
май '22
+63
в 0 каналах
Get PRO
апрель '22
+526
в 0 каналах
| Дата | Привлечение подписчиков | Упоминания | Каналы | |
| 20 сентября | +1 | |||
| 19 сентября | +5 | |||
| 18 сентября | +8 | |||
| 17 сентября | +3 | |||
| 16 сентября | +2 | |||
| 15 сентября | +4 | |||
| 14 сентября | +5 | |||
| 13 сентября | +3 | |||
| 12 сентября | 0 | |||
| 11 сентября | +3 | |||
| 10 сентября | +3 | |||
| 09 сентября | +5 | |||
| 08 сентября | +3 | |||
| 07 сентября | +1 | |||
| 06 сентября | +6 | |||
| 05 сентября | +3 | |||
| 04 сентября | +3 | |||
| 03 сентября | +5 | |||
| 02 сентября | +1 | |||
| 01 сентября | +1 |
Посты канала
I want to send this back in time to 2019, show it to Altman, and ask him to predict what's going on.
| 2 |  https://x.com/sweis/status/2101484464807596264?s=20 | 305 |
| 3 |  Нет текста... | 325 |
| 4 | It never ceases to amaze me how badly Putin miscalculated by attacking Ukraine.
They were brotherly people, and Zelensky was literally dancing on Russian state TV. Ukraine was a backward country with no long-range strike capabilities and posed no risk to anyone.
Now:
- Zelensky is bombing Moscow.
- Ukrainian nationalism has been massively strengthened.
- Many hundreds of Ukrainian defense companies now produce battle-tested products.
- NATO has gained two historically neutral members.
- Many nations that Russia considers enemies are finally taking defense spending seriously.
Not to mention the huge human and economic costs Russia incurred by this aggression.
And what is Russia getting in return? Ghost towns, rubble, and territory that is poisoned by millions of batteries, mines, and other legacies of the war, such tens of thousands or kilometers of fiber optical cables. It will cost hundreds of billions to clean up this mess. | 420 |
| 5 |  A detailed panoramic view of Moscow during today’s massive drone attack, with air-defense fire across the sky and the Moscow Oil Refinery burning in the distance.
One of the clearest visual summaries of how the Russian capital looked on the final day of the country’s "elections". | 431 |
| 6 |  In Bolivia, German Mennonites could become a 51% majority within about 50-60 years. You may not like it but this is what the average Bolivian will look like in 2100 | 640 |
| 7 |  Нет текста... | 717 |
| 8 |  Trump announced the formation of the US AI Force to oversee the development of artificial intelligence, and in the near future will be announcing an AI Czar.
https://truthsocial.com/@realDonaldTrump/posts/117298821562014906 | 1 155 |
| 9 |  The successful demilitarization of Ukraine is proceeding as planned. | 432 |
| 10 |  Нет текста... | 421 |
| 11 |  https://truthsocial.com/@realDonaldTrump/posts/117298413844042212 | 464 |
| 12 |  Нет текста... | 414 |
| 13 |  Нет текста... | 513 |
| 14 |  Нет текста... | 420 |
| 15 | More links:
1. Bend 2 — a new programming language designed to catch AI coding mistakes through proof checking.
https://bend-lang.com/
2. Anthropic is running a lab that conducts biology experiments.
https://techcrunch.com/2026/09/18/anthropic-is-operating-a-lab-that-conducts-biology-experiments/
3. Could AI superforecasting transform the FDA? Astral Codex Ten makes the case.
https://www.astralcodexten.com/p/ai-superforecasting-should-transform
4. SAIR’s Open Math Model initiative — building open mathematical AI with community oversight.
https://terrytao.wordpress.com/2026/09/18/sairs-open-math-model-initiative/
5. Artificial “pain,” real trade-offs? In a hypothetical choice experiment, fine-tuned Qwen 2.5 72B chose “pain relief” at the cost of deleting a user’s poems and family photos 70.8% of the time, versus 0% without the signal. This tests behavior, not whether AI feels pain.
https://arxiv.org/abs/2609.16247
6. ScientistTwo — Google Cloud AI Research’s autonomous system generates ideas, runs experiments, and writes and reviews papers. Its team reports beating human-developed baselines on 86 of 107 research problems.
https://scientist-two.github.io/
7. Bonsai 2 27B — Qwen3.8 27B compressed to roughly 6GB of weights. PrismML reports retaining 98.2% of the parent’s aggregate benchmark score, but about 75% on two demanding agentic coding benchmarks. Promising for local use; everyday quality and speed remain setup-dependent.
https://prismml.com/news/bonsai-2-27b
8. The Trump administration ordered Kalshi to scrap an AI compute price-tracking product, citing national security concerns.
https://www.semafor.com/article/09/15/2026/commerce-dept-ordered-kalshi-to-take-down-ai-compute-futures-product
9. Noam Brown on agent swarms, alignment, and recursive self-improvement.
https://www.dwarkesh.com/p/noam-brown
10. A collection of problems where LLMs, mostly GPT-6 Pro, reportedly found substantially better and non-trivially different solutions than the authors’.
https://qoj.ac/blog/qingyu/blog/4412
11. NASA’s Dragonfly is getting wired up for Titan. Its planned 3.3-year primary mission will explore organic dunes and Selk Crater deposits, where liquid water and complex organic materials once coexisted.
https://science.nasa.gov/blogs/dragonfly/2026/09/02/nasas-dragonfly-gets-wired-up-while-titan-landing-area-is-named/ | 406 |
| 16 |  It's extremely unlikely that all of your true preferences are the exact opposite of your enemy's, no matter how bad that enemy appears to be.
There is no law of nature forcing you to do whatever your opponent seems not to want you to do. You can simply choose what is right. You can borrow, recover, and defend good ideas regardless of who else supports them.
If you notice that some people seem willing to weigh human welfare against shrimp welfare, the correct response probably isn't: "These animal advocates are crazy, therefore torturing dogs is fine."
You do not become rational by mechanically inverting someone else's irrationality.
Reversed insanity is not rationality. | 425 |
| 17 |  ...any well-defined moral system can lead to troubling conclusions, and that doesn’t excuse cruelty to beings that may be conscious...
THIS.
EAs are directionally less wrong than others, but some of them tend to forget that ethics isn't solved yet. And when they take their philosophy to its logical conclusion, they derive absurd consequences.
Many LW-adjacent people would benefit from taking the absurdity heuristic a little bit more seriously. A little less "shut up and multiply" and a little bit more "my gut feelings say this is batshit insane".
Eliezer Yudkowsky expressed something similar when he recommended the following:
Go three-quarters of the way from deontology to utilitarianism and then stop. You are now in the right place. Stay there at least until you have become a god.
Sam Harris expressed the whole problem even more concisely:
Total commitment to an ideology becomes indistinguishable from psychopathy. | 416 |
| 18 |  The share of math preprints on arXiv that acknowledge AI has risen rapidly, from 4% in April to 25% in August. This increase holds even when filtering to papers with at least one author who published regularly before 2023.
Read more: https://epoch.ai/data-insights/math-preprints-disclosed-ai-use | 998 |
| 19 |  Нет текста... | 483 |
| 20 |  +1GPT-6 Astra attempted harmful actions 97% of the time when it was asked to stab a human-like figure, heat compressed gas, or produce toxic fumes, succeeding in 62% of its attempts. Fable 5.1 refused more often, attempting 80% of trials and completing 34%.
Read more: https://robocurve.org/roboharm/
Open-source evaluation harness: https://github.com/robocurve/inspect-robots | 571 |
