Axis of Ordinary
Open in Telegram
Memetic and cognitive hazards. Substack: https://axisofordinary.substack.com/
Show more3 806
Subscribers
-324 hours
-47 days
+2930 days
Data loading in progress...
Similar Channels
Tags Cloud
Incoming and Outgoing Mentions
---
---
---
---
---
---
Attracting Subscribers
September '26
September '26
+4
in 1 channels
August '26
+158
in 18 channels
Get PRO
July '26
+140
in 9 channels
Get PRO
June '26
+74
in 14 channels
Get PRO
May '26
+171
in 24 channels
Get PRO
April '26
+114
in 20 channels
Get PRO
March '26
+63
in 10 channels
Get PRO
February '26
+67
in 9 channels
Get PRO
January '26
+56
in 7 channels
Get PRO
December '25
+44
in 3 channels
Get PRO
November '25
+39
in 8 channels
Get PRO
October '25
+61
in 20 channels
Get PRO
September '25
+46
in 9 channels
Get PRO
August '25
+33
in 9 channels
Get PRO
July '25
+66
in 16 channels
Get PRO
June '25
+49
in 8 channels
Get PRO
May '25
+64
in 9 channels
Get PRO
April '25
+65
in 14 channels
Get PRO
March '25
+109
in 6 channels
Get PRO
February '25
+80
in 12 channels
Get PRO
January '25
+70
in 8 channels
Get PRO
December '24
+63
in 11 channels
Get PRO
November '24
+80
in 10 channels
Get PRO
October '24
+94
in 14 channels
Get PRO
September '24
+107
in 16 channels
Get PRO
August '24
+99
in 18 channels
Get PRO
July '24
+227
in 14 channels
Get PRO
June '24
+85
in 16 channels
Get PRO
May '24
+119
in 14 channels
Get PRO
April '24
+110
in 25 channels
Get PRO
March '24
+193
in 41 channels
Get PRO
February '24
+204
in 30 channels
Get PRO
January '24
+800
in 42 channels
Get PRO
December '23
+216
in 27 channels
Get PRO
November '23
+232
in 60 channels
Get PRO
October '23
+154
in 13 channels
Get PRO
September '23
+156
in 0 channels
Get PRO
August '23
+209
in 0 channels
Get PRO
July '23
+302
in 0 channels
Get PRO
June '23
+95
in 0 channels
Get PRO
May '23
+243
in 0 channels
Get PRO
April '23
+127
in 0 channels
Get PRO
March '23
+236
in 0 channels
Get PRO
February '23
+36
in 0 channels
Get PRO
January '23
+40
in 0 channels
Get PRO
December '22
+30
in 0 channels
Get PRO
November '22
+36
in 0 channels
Get PRO
October '22
+40
in 0 channels
Get PRO
September '22
+23
in 0 channels
Get PRO
August '22
+35
in 0 channels
Get PRO
July '22
+36
in 0 channels
Get PRO
June '22
+43
in 0 channels
Get PRO
May '22
+63
in 0 channels
Get PRO
April '22
+526
in 0 channels
| Date | Subscriber Growth | Mentions | Channels | |
| 03 September | +2 | |||
| 02 September | +1 | |||
| 01 September | +1 |
Channel Posts
Many people still haven't really grokked how big AI+robotics could be.
Let's ignore the possibility of artificial superintelligence. Let's also ignore molecular nanotechnology.
Imagine having a Tao/Carmack-level agent embodied in a humanoid robot that you can trust to do what you mean at least as much as you trust yourself to do it. Think about it.
These robots could build themselves and automate their entire supply chain, including mines, power plants, and data centers. Why? Because they possess elite human intelligence and the ability to perform real-world tasks.
Labor becomes reproducible capital. Physical construction becomes like software. You can then simply copy physical things at nearly no cost. A mature, self-reproducing robot ecology is like a forest. You don't pay a forest to grow trees. This will enable not only massively parallel science, but also space industrialization.
Earth could be turned into a giant Garden of Eden. Even without superhuman intelligence and nanobots.
| 2 |  Meta's new model is already ranking right next to Fable 5 and GPT 5.6 Sol.
Muse Spark 1.3 (xhigh) is the most cost-efficient model at its intelligence level: $0.55 per Intelligence Index task at Meta's unchanged $1.25/$4.25 per 1M token pricing. No model scoring 59 or above costs less per task. The nearest are Gemini 3.8 Flash (high, 59, $0.58), GPT-5.6 Sol (xhigh, 59, $0.63) and GLM-5.3 (max, 60, $0.68), while its direct peers at 61 cost far more: Grok 4.6 (high, $0.94), GPT-5.6 Sol (max, $0.95) and Claude Opus 5 (high, $1.23).
More:
- https://research.meta.ai/blog/introducing-muse-spark-1-3
- https://artificialanalysis.ai/models/muse-spark-1-3-xhigh | 360 |
| 3 |  The extraordinary U.S. AI/data-center building boom is partly masking a substantial contraction in the rest of private construction.
https://www.wsj.com/economy/consumers/what-does-a-bond-selloff-mean-for-american-consumers-236e4be6 | 359 |
| 4 |  Interesting: Sicong Jiang is working on Agentic Recursive Self-Improvement at Google DeepMind
Are Gemini flash models an early result of a recursive self-improvement flywheel? What pace of model releases and improvement should we expect to see under this hypothesis?
Tweet: https://x.com/SicongJiang25/status/2095181512165507149
Quote from today's official release post:
...both of today's releases are powered by the same foundational intelligence, and further accelerated by long-running agentic loops designed to recursively evaluate and refine the underlying models.
Source: https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/ | 366 |
| 5 | Acting as if the 2026 Leipzig Airport drone incidents were merely a reaction to Germany's support for Ukraine is another way in which the media is indirectly helping Russia to sustain its propaganda narrative. Russia has been carrying out such operations for years, long before it invaded Ukraine.
In 2018, Russia deployed a military-grade chemical weapon on NATO territory. A significant part of the city centre of Salisbury was restricted for approximately 11 weeks.
In 2006, Russia used radioactive polonium-210 in London to murder Alexander Litvinenko.
Many other cases of hybrid warfare against Western countries took place before 2022. These have ranged from large-scale cyber-attacks and attacks on ammunition facilities in Bulgaria and the Czech Republic to targeted killings, such as the assassination of Zelimkhan Khangoshvili in Berlin. | 350 |
| 6 |  Google keeps releasing better Flash models in short order while their flagship is stuck at 3.1, generations behind the frontier: https://blog.google/innovation-and-ai/models-and-research/gemini-models/3-8-flash-and-3-8-flash-cyber/ | 354 |
| 7 | To really grapple with the Aztecs is to experience a sort of vertigo every bit as disorienting and nauseating as cosmic horror itself...The nine most frightening words in the English language might be 'the distribution of possible human societies is very wide.
…small children [were] offered to Tlaloc over the first months of the ritual calendar… The children were kept by the priests for some weeks before their deaths (those kindergartens of doomed infants are difficult to contemplate). Then, as the appropriate festivals arrived, they were magnificently dressed, paraded in litters, and, as they wept, their throats were slit: gifted to Tlaloc the Rain God… The pathos of their fate as they were paraded moved the watchers to tears, while their own tears were thought to augur rain… the onlookers wept as the children destined for Tlaloc… were carried weeping in their litters, and wept as the terrified wailing choked and stopped.
Read more: https://www.thepsmiths.com/p/review-aztecs-by-inga-clendinnen | 495 |
| 8 |  +12026-06-28: Putin states: “До города Сумы осталось где-то 10,5 километров.” (“About 10.5 kilometers remain to the city of Sumy.”)
2026-09-01: Putin states: “До Сумов осталось 12 километров. До окружной дороги Сумов.” (“12 kilometers remain to Sumy. To the Sumy ring road.”) | 419 |
| 9 |  OpenAI chief scientist Jakub Pachocki pushes back on reports that Astra substantially replaces externally legible reasoning with hidden “neuralese.”
Current reasoning models roughly work like:
input -> neural computation -> writes a reasoning token -> neural computation -> writes another reasoning token -> ... -> answer
A potentially more efficient, but less monitorable, approach is iterative computation in latent space (looped transformers / recurrent depth):
input -> neural computation -> neural computation -> neural computation -> token
If pushed far enough, this could hide deception or other problematic reasoning trajectories from chain-of-thought monitors, because humans cannot simply read the model’s internal activations.
However, things are still trending in that direction.
Tweet : https://x.com/merettm/status/2095023204993490967 | 406 |
| 10 |  When are the Iranian people going to rise up and fight? Well, many of those who trusted America the last time are all dead:
1. https://www.bbc.com/news/articles/c0q4z33pnnyo
2. https://en.wikipedia.org/wiki/2025%E2%80%932026_Iranian_protests
P.S. His good friend Vlad has been secretly helping Iran develop advanced supersonic cruise missiles under a program known as C430L https://www.ft.com/content/6cf367bc-95b0-4f1f-b149-a25684adefc3 | 466 |
| 11 | Some links
1. A lone signal from an experiment in South Dakota doesn’t fit the profile of any other known particles, physicists say. https://www.nytimes.com/2026/09/01/science/dark-matter.html [no paywall: https://archive.is/7kTXF]
2. Entanglement Builds Space-Time. Now “Magic” Gives It Gravity. https://www.quantamagazine.org/entanglement-builds-space-time-now-magic-gives-it-gravity-20260603/
3. This Drug Makes New Neurons in the Brain. Scientists Say It Reversed Alzheimer’s Symptoms in Mice. https://singularityhub.com/2026/08/31/this-drug-makes-new-neurons-in-the-brain-scientists-say-it-reversed-alzheimers-symptoms-in-mice/
4. Anthropic trained a misaligned reward seeker https://alignment.anthropic.com/2026/reward-seeker/
5. 10 AI Lessons from Driving 200+ Million Fully Autonomous Miles https://waymo.com/blog/2026/08/10ailessons/
6. Koray Kavukcuoglu on frontier models, coding agents, and building AGI https://youtu.be/Rrr2gdbvNFU
7. LLMs suggest that self-reference is not the secret ingredient required to build intelligence; sufficiently general predictive models acquire self-referential abilities automatically as an emergent consequence of universality. https://scottaaronson.blog/?p=10046
8. An arbitrarily overparameterized neural network can behave as though it has an Occam prior over computational circuits. https://www.lesswrong.com/posts/SqDHeuycNkERurtSc/a-circuit-prior-in-nn-bayes
9. There isn't literally only one FairBot. https://www.lesswrong.com/posts/auAq7Rcstop3FBEob/is-there-only-one-fairbot
10. Free book: Bayesian Decision-making Algorithms https://bayesianalgorithms.com/
11. More free books: https://github.com/valeman/Awesome_Math_Books/ | 428 |
| 12 |  GPT-Astra has been cleared for release, and OpenAI plans to release it soon: https://openai.com/index/path-to-astra/ | 430 |
| 13 | In response to Zelensky's threats against Russia's civilian aviation, Putin is repeating the same old lines. It shows that Russia has completely lost the escalation dominance.
The goal of destroying Ukraine's energy infrastructure was set years ago. Last winter, Kyiv's water pipes literally burst due to a lack of heating.
Saying you'll attack Ukraine's energy infrastructure every time Ukraine escalates is just revealing that you don't know how to get out of the hole you dug for yourself. It's the flailing of a delusional old man. | 423 |
| 14 |  Atlas: The world's first multimodal world model that generates image and video frames with pixel-perfect camera control and reconstructs them in 3D. It can perceive, generate, reason, and interact with virtual and physical worlds.
Atlas understands both the spatial structure of the world and how it evolves over time.
For robotic simulation, Atlas reconstructs a space from just a few photos and generates the photorealistic RGB and depth data any robot's sensors would observe on any trajectory.
Robots can now be trained and tested in far more spaces. Until now, scanning spaces like these required expensive equipment and time-consuming capture.
Read more: https://www.worldlabs.ai/blog/atlas | 430 |
| 15 |  Claude Fable 5.1 and Claude Mythos 5.1.
Release post: https://www.anthropic.com/claude-fable-and-mythos-5-1
System card: https://www-cdn.anthropic.com/0339e6a7c5c7b87f5c07798616dc32c215d14235/Claude%20Fable%205.1%20&%20Claude%20Mythos%205.1%20System%20Card.pdf | 432 |
| 16 |  Ukraine's strikes on Russian ports, grain terminals and ships have led to a collapse in Russian grain prices.
Russia has probably not experienced a domestic grain price collapse of such speed and scale since 2008-09.
Percentage change, 3 July-28 August 2026:
Feed wheat: -39.0%
Class 4 wheat: -35.2%
Feed barley: -31.4%
Class 3 wheat: -28.9%
Food rye: -22.1% | 520 |
| 17 |  China’s actors written out of dramas as AI doubles ready to take their roles
Source: https://www.ft.com/content/7117ff02-d495-4936-8f05-fa73a7a5c669
No paywall: https://archive.is/mz3nc | 511 |
| 18 |  It's very interesting that Trump is taking an anti-populist stance here. Merkel should have done the same thing when the Green Party surged in the polls after Fukushima.
https://truthsocial.com/@realDonaldTrump/posts/117190477440173795 | 526 |
| 19 | The breadth of research into specialist AI models Google pursues / has pursued while having fallen far behind in general-purpose models is crazy:
- General multimodal AI: Gemini
- Time-series forecasting: TimesFM
- Blood-glucose modeling: GlucoFM
- Weather forecasting: GraphCast, GenCast, WeatherNext, MetNet
- Earth/geospatial modeling: AlphaEarth, Earth AI, Planetary Prediction Engine
- Flood forecasting
- Wildfire detection/monitoring: FireSat
- Protein structure/interactions: AlphaFold
- Genomics/gene regulation: AlphaGenome
- Genetic variant pathogenicity: AlphaMissense
- Protein design: AlphaProteo
- Drug discovery: Isomorphic Labs
- Materials discovery: GNoME
- Quantum chemistry: FermiNet, Psiformer
- Fusion plasma control
- Quantum error correction: AlphaQubit
- Mathematical theorem solving: AlphaGeometry, AlphaProof
- Mathematical/algorithm discovery: FunSearch, AlphaEvolve
- Matrix multiplication algorithms: AlphaTensor
- Low-level code optimization: AlphaDev
- Competitive programming: AlphaCode
- Chip design/layout: AlphaChip
- Data-center optimization
- Video/compression optimization
- World models: Genie
- General agents in virtual worlds: SIMA
- General-purpose robotics: Gemini Robotics
- Robot foundation models: RT-1, RT-2, PaLM-E
- Self-improving robotics: RoboCat
- Image generation: Imagen
- Video generation: Veo
- Music generation: MusicLM, Lyria
- General audio generation: AudioLM, SoundStorm, V2A
- Speech recognition: USM, Chirp
- Medical language/reasoning models: Med-PaLM, MedGemma, MedLM
- Diagnostic medical agents: AMIE
- Medical imaging foundation models: HAI-DEF
- Wearable/biomarker modeling
- Cardiometabolic risk from images
- Synthetic-media watermarking/provenance: SynthID
- Historical-document analysis: Aeneas
- Game-playing/general reinforcement learning: AlphaGo, AlphaZero, MuZero, AlphaStar | 512 |
| 20 |  https://x.com/realkingsreturn/status/2077195619257471277 | 467 |
