UC Berkley used OpenEvolve to discover algorithms across diverse domains that “outperform SOTA human designs”, achieving up to 5.0× runtime improvements or 50% cost reductions.

Conclusion:

One of the most profound implication is the potential for a virtuous cycle. We can use an ADRS to improve itself. As recent work has shown, models can learn to refine their own reasoning, debug their code, and discover more effective strategies.

Paper: https://arxiv.org/abs/2510.06189

Links for 2025-10-11

AI

1. Large Language Models Achieve Gold Medal Performance at the International Olympiad on Astronomy & Astrophysics (IOAA) https://arxiv.org/abs/2510.05016

2. Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models — “The model writes, reflects, and rewrites its own prompt over and over until it becomes a self-improving system” https://www.arxiv.org/abs/2510.04618

3. Nvidia researchers boost LLMs reasoning skills by getting them to ‘think’ during pre-training https://venturebeat.com/ai/nvidia-researchers-boost-llms-reasoning-skills-by-getting-them-to-think

4. h1: Bootstrapping LLMs to Reason over Longer Horizons via Reinforcement Learning https://arxiv.org/abs/2510.07312

5. Code World Models for General Game Playing https://arxiv.org/abs/2510.04542

6. Towards a Typology of Strange LLM Chains-of-Thought https://www.lesswrong.com/posts/qgvSMwRrdqoDMJJnD/towards-a-typology-of-strange-llm-chains-of-thought

7. Researchers working at the frontier labs are genuinely convinced that transformative AI is coming. https://www.transformernews.ai/p/were-all-behind-the-curve-ai-bubble-crash-risk

8. Talking about the Future of AI in Law with David Wakeling https://artificialinvestment.substack.com/p/interview-with-david-wakeling-on

9. My Favorite Streamer is an LLM: Discovering, Bonding, and Co-Creating in AI VTuber Fandom https://arxiv.org/abs/2509.10427

10. Moloch’s Bargain: Emergent Misalignment When LLMs Compete for Audiences https://arxiv.org/abs/2510.06105

11. How a 23-year-old former OpenAI researcher turned a viral AI prophecy into profit, with a $1.5 billion hedge fund and outsize influence from Silicon Valley to D.C. https://fortune.com/2025/10/08/leopold-aschenbrenner-openai-ftx-1-5-billion-hedge-fund-situational-awareness/

12. A small number of samples can poison LLMs of any size https://www.anthropic.com/research/small-samples-poison

13. How well can large language models predict the future? https://forecastingresearch.substack.com/p/ai-llm-forecasting-model-forecastbench-benchmark

Science and Engineering

1. UMass Engineers Create First Artificial Neurons That Could Directly Communicate With Living Cells https://www.umass.edu/news/article/umass-engineers-create-first-artificial-neurons-could-directly-communicate-living

2. A cGAS-mediated mechanism in naked mole-rats potentiates DNA repair and delays aging https://www.science.org/doi/10.1126/science.adp5056

3. Scientists Just Reversed Alzheimer’s in Mice. Could They Do It in Humans? https://gizmodo.com/scientists-just-reversed-alzheimers-in-mice-could-they-do-it-in-humans-2000669049

4. Programmable Aptamer-Embedded Circular RNAs for Targeted Antigen-Presenting Cells Immunotherapy https://www.biorxiv.org/content/10.1101/2025.09.28.679023v1

5. Why China Built 162 Square Miles of Solar Panels on the World’s Highest Plateau https://www.nytimes.com/2025/10/10/business/china-solar-tibetan-plateau.html [no paywall: https://archive.is/vuOYu]

6. The government’s own data rebuts Trump’s claims about wind and solar prices https://www.politico.com/news/2025/10/07/green-electricity-costs-cheap-trump-00594123

Total trade war?

1. “China has asserted sweeping control over the entire global semiconductor supply chain, putting export license requirements on all rare earths used to manufacture advanced chips. If enforced aggressively, this policy could mean “lights out” for the US AI boom, and likely lead to a recession/economic crisis in the US in the short term.” https://x.com/deanwball/status/1976260051351343195

2. October 9 2025 Ministry of Commerce/GAC Export Control actions https://sinocism.notion.site/October-9-2025-Ministry-of-Commerce-GAC-Export-Control-actions-28784ece41d78059ba33cb4cca154bf4

This is Charlie Gee

One of the world’s most respected and renowned stonemason

His work in performing restoration works on Cathedrals throughout Europe is some of the best

A skill set that is becoming rarer to find and master in a more modern world

He’s only 23 years old

A rock lath master at work in the 1940s USA

Chart from the Federal Reserve Bank of Dallas: https://www.dallasfed.org/research/economics/2025/0624

With the Russian summer offensive a complete failure and no serious autumn offensive in sight, let's review what Russia has achieved in 3 years, 7 months, and 19 days of the 3-day special military operation:

- Constant Ukrainian attacks and assassinations throughout Russia, accompanied by a worsening fuel crisis.
- The NATO-Russia border has expanded by 1,340 km as the formerly neutral countries of Sweden and Finland have joined NATO.
- Pushing pacifist, pro-Russian countries like Germany to mass-produce autonomous killer drones and forcing German defense enterprises, such as Rheinmetall, to supply Ukraine instead of honoring their planned billion-euro contracts with Russia.
- Loss of the majority of the European export market, resulting in dependence on China.
- The American president and NATO Secretary General are calling Russia a paper tiger.
- Millions of Muslim immigrants from former Soviet satellite states are being brought in to alleviate labor shortages.
- Putin had to kiss the Quran to appease the Muslim minority fighting for Russia.
- The Black Sea Fleet had to flee Crimea after losing more than 33-40% of its assets, only to be attacked in Novorossiysk as well.
- At least 7,032 Russian officers have been killed, as confirmed by public Russian obituaries and gravesites. That's more than twice the number of people killed on 9/11, but they are all military leaders.

What's the best-case outlook for Russia? Capturing a completely destroyed Donbass while Ukraine starts the mass-production of cruise and ballistic missiles, becoming a greater danger to Russia than it has ever been.

Google is well known for using AI to design its in-house processors (TPUs). Their internal model, AlphaChip, dramatically accelerated the design cycle and unlocked superhuman chip layouts.

Now, OpenAI is doing the same: https://youtu.be/qqAbVTFnfk8?si=dSizokgkDVrxYmUk&t=366

Quote:

We've been able to apply our own models to designing this chip... We've been able to get massive area reductions. You take components that humans have already optimized and just pour compute into it, and the model comes up with its own optimizations.

We're at the point now where I don't think any of the optimizations we have are ones that human designers couldn't have come up with. Usually our experts take a look at it later and say, yeah, this was on my list, but it was a list of 20 things that would have taken them another month to get to.

We were coming up on a deadline and we were running optimizations. We had a choice of, do we actually take a look at what those optimizations were or do we just keep going until the deadline and then take a look after? And we decided, of course, you got to just keep going.

This video looks fake but is indeed real: Starship touchdown

Google Veo is getting a major upgrade.

Read more: https://blog.google/technology/ai/veo-updates-flow/

Google and Yale scientists have trained an LLM that has generated a novel hypothesis about cancer cellular behavior. This prediction was confirmed multiple times in vitro.

What made this prediction so exciting was that it was a novel idea. Although CK2 has been implicated in many cellular functions, including as a modulator of the immune system, inhibiting CK2 via silmitasertib has not been reported in the literature to explicitly enhance MHC-I expression or antigen presentation. This highlights that the model was generating a new, testable hypothesis, and not just repeating known facts.


The model that generated this prediction is a 27B-parameter LLM based on the Google Gemma open source models, and trained on a corpus comprising >1B tokens of transcriptomic data, biological text, and metadata. Quite remarkable that a small (just 27B) LLM trained on specialized data is able to make novel scientific discoveries.

Read more: https://blog.google/technology/ai/google-gemma-ai-cancer-therapy-discovery/

Unplanned shutdowns at several key refineries due to Ukrainian strikes reduced fuel exports and led to an increase in crude oil shipments. Exports of petroleum products from Russia by sea in September fell by 17.1% compared to August — to 7.58 million tons.

The impact from Ukrainian drone strikes will suppress Russia’s refinery processing rates until at least mid-2026, the International Energy Agency says.

The increasingly widespread and significant Ukrainian drone campaign against Russian oil refineries and infrastructure' has so far cut the nation’s crude processing by an estimated 500,000 barrels a day.

Sources:
- https://www.reuters.com/business/energy/russias-seaborne-fuel-exports-sank-17-september-after-drone-attacks-2025-10-13/
- https://www.bloomberg.com/news/articles/2025-10-14/iea-sees-drone-strikes-weighing-on-russia-oil-processing-till-mid-2026

Some of the Russian energy infrastructure hit in the past few days:

- In Volgograd region, the 500 kV "Balashovskaya" substation near Uryupinsk was hit

- The Saratov refinery

- The Bashneft oil refinery in Ufa

- Feodosia fuel depot

- The 490 MW Tavri power plant

- The 117 MW Saki Thermal Power Plant and two electrical substations (220 kV and 330 kV)

- The powerplant in Russian Belgorod

- The oil depot in Matveev Kurgan, Rostov region

- The Korobkovsky gas processing plant and the LPDS “Yefimovka” (a linear oil pumping station) in the Volgograd region

Links for 2025-10-16 [Part 1]

AI

1. Recursive Language Models (RLMs): On the OOLONG benchmark, RLMs with GPT-5-mini outperforms GPT-5 by over 110% gains (more than double!) on 132k-token sequences and is cheaper to query on average. On the BrowseComp-Plus benchmark, RLMs with GPT-5 can take in 10M+ tokens as their “prompt” and answer highly compositional queries without degradation and even better than explicit indexing/retrieval. https://alexzhang13.github.io/blog/2025/rlm/

2. The Art of Scaling Reinforcement Learning Compute for LLMs https://arxiv.org/abs/2510.13786

3. AI is apparently already accelerating science. Measuring academic publications of authors: “we find that productivity among GenAI users rose by 15 percent in 2023 relative to non-users and further increased to 36 percent in 2024” and the quality of publications also went up. https://arxiv.org/abs/2510.02408

4. Physics-informed AI excels at large-scale discovery of new materials https://phys.org/news/2025-10-physics-ai-excels-large-scale.html

5. ChatGPT-Like AI Unveils 1,300 Regions in the Mouse Brain—Some Uncharted https://singularityhub.com/2025/10/16/chatgpt-like-ai-unveils-1300-regions-in-the-mouse-brain-some-uncharted/

6. Bringing AI to the next generation of fusion energy https://deepmind.google/discover/blog/bringing-ai-to-the-next-generation-of-fusion-energy/

7. OpenAI hires black hole physicist in broader science push https://www.axios.com/2025/10/16/openai-science-black-hole-physicist

8. SR-Scientist: Scientific Equation Discovery With Agentic AI https://arxiv.org/abs/2510.11661

9. Agent Learning via Early Experience https://arxiv.org/abs/2510.08558

10. A skull-stripping model trained using purely synthetic data, no real brains or labels required. It generalizes across species, modalities, and pathologies out-of-the-box without additional fine-tuning. https://arxiv.org/abs/2505.07159

11. DeepMMSearch-R1: Empowering Multimodal LLMs in Multimodal Web Search https://arxiv.org/abs/2510.12801

12. Meta-Awareness Enhances Reasoning Models: Self-Alignment Reinforcement Learning https://arxiv.org/abs/2510.03259

13. StreamingVLM: Real-Time Understanding for Infinite Video Streams https://arxiv.org/abs/2510.09608

14. Pretraining Large Language Models with NVFP4 https://www.arxiv.org/abs/2509.25149

15. Training-Free Group Relative Policy Optimization https://arxiv.org/abs/2510.08191

16. Not All Bits Are Equal: Scale-Dependent Memory Optimization Strategies for Reasoning Models https://arxiv.org/abs/2510.10964

17. CMT-Benchmark: A Benchmark for Condensed Matter Theory Built by Expert Researchers https://arxiv.org/abs/2510.05228

18. Is there some meaningful sense in which, say, the unusual scientific or economic productivity of Isaac Newton or James Watt, can be described in terms of the “time horizon” of their minds? https://www.lesswrong.com/posts/PzLSuaT6WGLQGJJJD/the-length-of-horizons

19. gpt5-pro is superhuman at literature search https://x.com/SebastienBubeck/status/1977181716457701775

20. GPT 5 Pro is extremely good in identifying serious gaps in published papers. https://x.com/PI010101/status/1977117411603366363

21. Inside Google’s AI turnaround: AI Mode, AI Overviews, and vision for AI-powered search https://www.youtube.com/watch?v=kOnsqqVbIeY

22. A repo with full code and instructions to walk through tuning your own GPT-2 level LLM in an afternoon, including all of the steps. https://github.com/karpathy/nanochat/discussions/1

23. A comprehensive, hands-on tutorial on the most recent advancements in robotics https://arxiv.org/abs/2510.12403

24. You can make ChatGPT 2x as creative with one sentence. https://www.verbalized-sampling.com/

25. Theory of Mind differentially predicts performance during collaboration with AI but not when working alone. https://osf.io/preprints/psyarxiv/vbkmt_v1

Links for 2025-10-16 [Part 2]

AI

26. “How I Became a 5x Engineer with Claude Code” https://www.lesswrong.com/posts/pcFSo5hEzzGTZEXEd/how-i-became-a-5x-engineer-with-claude-code

27. OpenAI and Broadcom announce strategic collaboration to deploy 10 gigawatts of OpenAI-designed AI accelerators https://openai.com/index/openai-and-broadcom-announce-strategic-collaboration/

28. Meta: “Located in El Paso, Texas, this will be our 29th data center, and will help us to deliver top-tier AI models as we work to fulfill the possibilities of AI and build toward superintelligence.” https://about.fb.com/news/2025/10/metas-new-ai-optimized-data-center-el-paso/

29. Coco Robotics taps UCLA professor to lead new physical AI research lab https://techcrunch.com/2025/10/14/coco-robotics-taps-ucla-professor-to-lead-new-physical-ai-research-lab/

30. OpenAI is projecting unprecedented revenue growth https://epochai.substack.com/p/openai-is-projecting-unprecedented

Neurotech

1. Scientists create nanofluidic chip with ‘brain-like’ memory pathways https://phys.org/news/2025-10-scientists-nanofluidic-chip-brain-memory.html

2. These Super-Efficient, Artificial Neurons Do Not Use Electrons https://spectrum.ieee.org/these-artificial-neurons-use-ions-rather-than-electrons

3. In a First, Artificial Neurons Talk Directly to Living Cells. Bacterial nanowires make memristors operate at cellular voltages. https://spectrum.ieee.org/artificial-neuron

4. Researchers reveal connectome of the male fruit fly central nervous system https://www.janelia.org/news/researchers-reveal-connectome-of-the-male-fruit-fly-central-nervous-system

Science

1. A DNA search engine https://ethz.ch/en/news-and-events/eth-news/news/2025/10/a-dna-search-engine.html

2. Harvard and Peking University scientists have built new tools that can edit the DNA inside mitochondria, long unreachable by CRISPR. This could soon allow treatments for at least 150 diseases linked to faulty mitochondria, from blindness to heart failure. https://www.nature.com/articles/d41586-025-03307-x [no paywall: https://archive.is/e3eax]

3. A Nanoparticle Drug Triggered the Brain to Rapidly Flush Out Toxic Alzheimer’s Proteins in Mice https://singularityhub.com/2025/10/13/a-nanoparticle-drug-triggered-the-brain-to-rapidly-flush-out-toxic-alzheimers-proteins-in-mice/

4. A US man’s failure to develop Alzheimer’s disease is baffling researchers and challenging scientific understanding of the condition. https://www.nytimes.com/2025/10/07/health/alzheimers-gene-mutation.html [no paywall: https://archive.is/Oedef]

5. More Evidence Emerges That One of Saturn’s Moons Could Harbor Life https://www.wired.com/story/more-evidence-emerges-that-one-of-saturns-moons-could-harbor-life/ [no paywall: https://archive.is/LMYKa]

Politics

1. JPMorgan will make direct equity and venture capital investments of up to $10 billion into US companies deemed critical to national security. — “It has become painfully clear that the United States has allowed itself to become too reliant on unreliable sources of critical minerals, products and manufacturing – all of which are essential for our national security” https://www.jpmorganchase.com/newsroom/press-releases/2025/jpmc-security-resiliency-initiative

Despite constant claims to the contrary, AI is generally undervalued. That doesn't mean there won't be temporary market upheavals where valuations crash massively. Just remember how irrationally the market reacted to DeepSeek, when anyone who followed AI closely was unfazed.[1][2] Nvidia fell ~17% in a day, for no good reason at all. But such a crisis will only be of a temporary nature.

Very few insiders doubt that AI will be transformative. What distinguishes informed skeptics from believers these days is whether the disruptions will start in 2 or 10 years.

[1] https://t.iss.one/axisofordinary/6920
[2] https://t.iss.one/axisofordinary/6925