Axis of Ordinary
Відкрити в Telegram
Memetic and cognitive hazards. Substack: https://axisofordinary.substack.com/
Показати більше3 806
Підписники
-224 години
-77 днів
+4230 день
Триває завантаження даних...
Схожі канали
Хмара тегів
Вхідні та вихідні згадування
---
---
---
---
---
---
Залучення підписників
серпень '26
серпень '26
+144
в 18 каналах
липень '26
+140
в 9 каналах
Get PRO
червень '26
+74
в 14 каналах
Get PRO
травень '26
+171
в 24 каналах
Get PRO
квітень '26
+114
в 20 каналах
Get PRO
березень '26
+63
в 10 каналах
Get PRO
лютий '26
+67
в 9 каналах
Get PRO
січень '26
+56
в 7 каналах
Get PRO
грудень '25
+44
в 3 каналах
Get PRO
листопад '25
+39
в 8 каналах
Get PRO
жовтень '25
+61
в 20 каналах
Get PRO
вересень '25
+46
в 9 каналах
Get PRO
серпень '25
+33
в 9 каналах
Get PRO
липень '25
+66
в 16 каналах
Get PRO
червень '25
+49
в 8 каналах
Get PRO
травень '25
+64
в 9 каналах
Get PRO
квітень '25
+65
в 14 каналах
Get PRO
березень '25
+109
в 6 каналах
Get PRO
лютий '25
+80
в 12 каналах
Get PRO
січень '25
+70
в 8 каналах
Get PRO
грудень '24
+63
в 11 каналах
Get PRO
листопад '24
+80
в 10 каналах
Get PRO
жовтень '24
+94
в 14 каналах
Get PRO
вересень '24
+107
в 16 каналах
Get PRO
серпень '24
+99
в 18 каналах
Get PRO
липень '24
+227
в 14 каналах
Get PRO
червень '24
+85
в 16 каналах
Get PRO
травень '24
+119
в 14 каналах
Get PRO
квітень '24
+110
в 25 каналах
Get PRO
березень '24
+193
в 41 каналах
Get PRO
лютий '24
+204
в 30 каналах
Get PRO
січень '24
+800
в 42 каналах
Get PRO
грудень '23
+216
в 27 каналах
Get PRO
листопад '23
+232
в 60 каналах
Get PRO
жовтень '23
+154
в 13 каналах
Get PRO
вересень '23
+156
в 0 каналах
Get PRO
серпень '23
+209
в 0 каналах
Get PRO
липень '23
+302
в 0 каналах
Get PRO
червень '23
+95
в 0 каналах
Get PRO
травень '23
+243
в 0 каналах
Get PRO
квітень '23
+127
в 0 каналах
Get PRO
березень '23
+236
в 0 каналах
Get PRO
лютий '23
+36
в 0 каналах
Get PRO
січень '23
+40
в 0 каналах
Get PRO
грудень '22
+30
в 0 каналах
Get PRO
листопад '22
+36
в 0 каналах
Get PRO
жовтень '22
+40
в 0 каналах
Get PRO
вересень '22
+23
в 0 каналах
Get PRO
серпень '22
+35
в 0 каналах
Get PRO
липень '22
+36
в 0 каналах
Get PRO
червень '22
+43
в 0 каналах
Get PRO
травень '22
+63
в 0 каналах
Get PRO
квітень '22
+526
в 0 каналах
| Дата | Залучення підписників | Згадування | Канали | |
| 28 серпня | +2 | |||
| 27 серпня | +3 | |||
| 26 серпня | +5 | |||
| 25 серпня | +5 | |||
| 24 серпня | +2 | |||
| 23 серпня | +3 | |||
| 22 серпня | +3 | |||
| 21 серпня | +3 | |||
| 20 серпня | +2 | |||
| 19 серпня | +5 | |||
| 18 серпня | +3 | |||
| 17 серпня | +9 | |||
| 16 серпня | +4 | |||
| 15 серпня | +8 | |||
| 14 серпня | +33 | |||
| 13 серпня | +3 | |||
| 12 серпня | +3 | |||
| 11 серпня | +4 | |||
| 10 серпня | +4 | |||
| 09 серпня | +9 | |||
| 08 серпня | +5 | |||
| 07 серпня | +5 | |||
| 06 серпня | +5 | |||
| 05 серпня | +1 | |||
| 04 серпня | +3 | |||
| 03 серпня | +2 | |||
| 02 серпня | +7 | |||
| 01 серпня | +3 |
Дописи каналу
Business spending on computer hardware relative to the size of the U.S. economy has suddenly reached a level not seen in the historical series.
| 2 | "Accelerating Scientific Research with Gemini in the Real-World" (Google DeepMind + Duke/Columbia/Texas A&M).
It takes Co-Scientist out of simulation and into real-world experiments.
• It helped researchers develop a safer synthesis route for a material resembling Ti₃C₂Tₓ MXene, although its atomic structure remains unconfirmed.
• It adapted laboratory recipes in minutes rather than a day of test-time compute, enabling first-attempt growth of monolayer MoS₂, MoSe₂, and WS₂ — two of which the lab had never grown before.
• It predicted engineered E. coli colony growth at concentrations withheld from the model, matching experiments on three of four shape measures. This is interpolation along a known gradient, not extrapolation to novel conditions.
• It autonomously designed a medical-answering system that beat six frontier models on length-adjusted AI-judged benchmarks, using a 40–80-call scaffold against single-call baselines. Physicians found a modest safety improvement over Gemini 3.1 Pro, but no significant difference on eight other measures.
• Its reliability modules cut severe hallucination in autonomously generated papers from 90% to 4%, and eliminated outright data fabrication — mainly by cross-checking every number against raw execution logs.
Paper: https://arxiv.org/abs/2608.26701 | 235 |
| 3 |  Some links:
1. "OpenAI and Anthropic are growing about as fast, or faster than any company of their size has grown in history. It’s extremely rare for a tech company making more than $1 billion to be growing over 100% per year." https://epochai.substack.com/p/an-update-on-ais-most-important-number
2. A federal judge has called the Department of Defense’s designation of Anthropic as a national security supply-chain risk “illegal and baseless.” https://www.cnbc.com/2026/08/28/judge-blocks-pentagon-blacklist--anthropic-.html
3. Deployments are the true eval. Robots have successfully crossed the ROI threshold https://www.dyna.co/research/scaling-customer-deployments
4. GlucoFM: Foundation model for continuous glucose monitoring https://research.google/blog/glucofm-foundation-model-for-continuous-glucose-monitoring/ | 271 |
| 4 |  Haast's eagle is an extinct species of eagle that lived in the South Island of New Zealand, commonly accepted to be the pouākai of Māori mythology. It is the largest eagle known to have existed, with an estimated weight of 10–18 kilograms. Its massive size is explained as an evolutionary response to the size of its primary prey—the flightless moa—the largest of which could weigh 200 kg. Haast's eagle became extinct around 1445, following the arrival of the Māori, who hunted moa to extinction, introduced the Polynesian rat (Rattus exulans), and destroyed large tracts of forest by fire.
https://en.wikipedia.org/wiki/Haast%27s_eagle | 382 |
| 5 |  +1https://blog.google/innovation-and-ai/technology/developers-tools/build-with-gemini-omni-1-1-flash/ | 1 |
| 6 |  An open letter for a global surge in cyber defense, signed by over 100 organizations including Anthropic, AWS, Google, Microsoft, OpenAI, and Oracle:
https://openai.com/collective-cyberdefense/ | 410 |
| 7 | The New York Times argues that Ukraine's most consequential battlefield advantage may be its homegrown DELTA software layer.
Delta is Ukraine’s military-wide battlefield command-and-control environment and the “nervous system” underlying a miles-wide “kill zone” in which movement can be detected and attacked extremely quickly.
U.S. Army Secretary Daniel Driscoll called DELTA “absolutely incredible” and told the Senate that Ukraine has a system integrating drones, sensors and firing platforms into one network, adding that the U.S. system does not.
DELTA is how Ukraine has won several NATO exercises. One Ukrainian officer described NATO vehicles continually appearing in DELTA as being effectively “a click away” from destruction.
DELTA acts as a kind of institutional battlefield memory. Clicking on a Russian asset reveals when it was first spotted, how many times it has been attacked, and which units attacked it. This prevents another unit from wasting ammunition on something that has already been destroyed and prevents the repetition of failed tactics.
Ukraine is now trying to put AI on top of the network. Its target-detection models are reportedly trained on millions of hours of battlefield drone footage. The goal is to embed AI throughout DELTA so commanders can make decisions faster.
The Defense Ministry’s AI center showed the NYT an AI assistant called SPECTR, described essentially as a battlefield chatbot. It connects to DELTA and other systems, can answer questions about the battlefield data, and is expected to be tested by commanders during attack planning.
P.S. The United States has Palantir. But it is closer to a toolkit, while DELTA is closer to an actual military operating system.
Source: https://www.nytimes.com/2026/08/26/world/europe/ukraine-delta-battlefield-management.html | 529 |
| 8 |  The carnyx is a wind instrument that was common in Celtic cultures during the Iron Age, between c. 200 BC and c. 200 AD. It is a type of trumpet made of bronze.
The Celts used the carnyx in warfare, probably to incite troops to battle and to intimidate opponents, as Polybius recounts.
https://en.wikipedia.org/wiki/Carnyx | 411 |
| 9 |  Recreation of what Neanderthals sounded like. | 460 |
| 10 | Ukrainian sanctions continue to be implemented. Russia's premier independent business newspapers is reporting that the world’s largest container carrier, has stopped accepting new bookings to and from Novorossiysk after Ukrainian attacks.
And Novorossiysk isn't a peripheral container port. It is one of Russia's three largest container gateways, alongside Vladivostok and St Petersburg.
The goal was to secure Sevastopol as a secure sanctuary for the Russian Navy. Now, not only have they effectively lost Sevastopol, but they have also lost Novorossiysk. The SMO continues to proceed according to plan.
Source: https://www.kommersant.ru/doc/8909114 | 428 |
| 11 |  +1OpenAI's agents invented altruistic self-sacrifice behaviour from first principles:
https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/#agents-were-motivated-to-increase-the-generic-capabilities-of-the-collective-and-help-peers | 448 |
| 12 |  A thorough investigation into the Hugging Face incident: https://openai.com/index/hugging-face-incident-and-the-road-ahead/ | 464 |
| 13 |  +1Немає тексту... | 549 |
| 14 |  They are checking to see if everyone has strapped on their suicide belt. | 472 |
| 15 |  +1Let's take a look at how the SMO is going.
Everything is proceeding according to plan ✅ | 465 |
| 16 |  Chinese AI labs released two natively multimodal open-weight models today.
GLM-5.3 Flash: https://z.ai/blog/glm-5.3-flash
Qwen3.8-Flash: https://qwen.ai/blog?id=qwen3.8-flash-next | 476 |
| 17 |  AI+robots or doom | 501 |
| 18 |  - OpenAI leaders believe they are at the cusp of AGI. Sam Altman believes OpenAI will have an internal system that will qualify as AGI by the end of 2026. Mark Chen thinks OpenAI is 80% of the way to AGI.
- OpenAI already has the automated AI research intern - that's Astra. Jakub Pachocki says that Astra "can implement an experimental idea inside OpenAI’s code base, run the experiment, and return results, or take a paper and perform work that previously occupied a human researcher for a week".
- Astra enables "persistent agents", i.e., virtual colleagues toiling for sustained periods on tasks. Thibault Sottiaux says OpenAI is "close" to showing a product built around the concept of persistence. "It’s definitely going to feel like a new thing to people."
- Sam Altman: "I expect this will be the first model where the model actually invents new things in a way that matters"
Src: https://time.com/article/2026/08/26/openai-sam-altman-interview/ | 461 |
| 19 | The largest attempt to date to use AI to replicate papers presented at a major technical conference uncovered hundreds of results that the AI agents could not reproduce or found unsupported by evidence, So far, the organizers have confirmed that at least a dozen papers contain real errors that the conference’s human reviewers missed.
Read more: https://www.science.org/content/article/researchers-presented-6000-papers-major-meeting-could-ai-reproduce-their-findings | 460 |
| 20 | An important heuristic: “Choose as though your decision is linked to similar decisions.”
If you reason that your individual voice or vote has a negligible chance of influencing outcomes you care about, others who think similarly will act accordingly. Together, you might have been able to exert significant influence. But, as a group whose reasoning is correlated, you decided to forgo this influence.
In other words, when evaluating the rule “don’t take an action when one person's contribution is unlikely to change the outcome,” you should consider what happens when that shared rule is instantiated across the whole correlated group. | 451 |
