Image created with gemini-2.5-flash-image with claude-sonnet-4-5. Image prompt: Photorealistic architectural photography of six freestanding Ionic limestone columns on a university quad with a completed classical entablature spanning the top, the word ETHICS carved in centered Roman serif letters in the frieze alongside a bas-relief of balanced scales of justice, warm beige limestone texture, late afternoon golden hour light, red brick buildings and green lawn in background, wide landscape composition with crisp realism and soft shadows.
Agents Rule of Two: A Practical Approach to AI Agent Security https://ai.meta.com/blog/practical-ai-agent-security/
A new security agent called Aardvark:”” / X https://x.com/sama/status/1984002552158154905
Now in private beta: Aardvark, an agent that finds and fixes security bugs using GPT-5. https://x.com/OpenAI/status/1983956431360659467
Accumulating Context Changes the Beliefs of Language Models https://arxiv.org/pdf/2511.01805
Bullying is Not Innovation https://www.perplexity.ai/hub/blog/bullying-is-not-innovation
Emergent introspective awareness in large language models \ Anthropic https://www.anthropic.com/research/introspection
Emergent introspective awareness in LLMs Anthropic used concept injection to test whether LLMs can introspect on their internal states. They found that: – Claude Opus 4.1 and 4 detected injected concepts with 20% success at optimal layers. They distinguished internal “”thoughts”” https://x.com/TheTuringPost/status/1986220265253314895
What’s up with Anthropic predicting AGI by early 2027? — LessWrong https://www.lesswrong.com/posts/gabPgK9e83QrmcvbK/what-s-up-with-anthropic-predicting-agi-by-early-2027-1
** HIGHLY CONFIDENTIAL **
13 Videotaped Deposition of ILYA SUTSKEVERgov.uscourts.cand.433688.340.1.pdf https://storage.courtlistener.com/recap/gov.uscourts.cand.433688/gov.uscourts.cand.433688.340.1.pdf
Lots of RTs on my Helen Toner interview today. Over a year later, we hear Ilya’s side of this story. The Brockman beef was unexpected (Ilya wrote a memo on him too), as was the failed Anthropic merger. Interesting that most screenshots in Ilya’s 52 page memo were sourced from”” / X https://x.com/bilawalsidhu/status/1985253917106827682
IndQA is a new benchmark designed to evaluate how well AI systems understand culture, context and history to answer questions that matter to people in India. With 2278 questions created in partnership with 250+ experts, IndQA dives deep into reasoning about everyday life,”” / X https://x.com/snsf/status/1985719755551158754
Meta, Google, Apple – they’re all building AI replicas that capture your face, expressions, movements, personality. This goes way beyond Face ID. They’re basically creating a version of you that knows you better than you know yourself. The fidelity is remarkable too. We went https://x.com/bilawalsidhu/status/1985398951407722901
We wanted to share more information about Gemma in AI Studio: First, to clarify the distinction between our AI products. Our Gemma models are a family of open models built specifically for the developer and research community. They are not meant for factual assistance or for”” / X https://x.com/NewsFromGoogle/status/1984412221531885853
China issues 50% electricity subsidies for datacenters. Very clever. Their industrial energy costs are already (generally) below the US and won’t spike due to such trifle as datacenters, but their chips are far less efficient. With this they get to > Hopper levels of FLOPs/$. https://x.com/teortaxesTex/status/1985540154065318157
Meta estimates that it earns 10% of its revenue from scams, report says | TechCrunch https://techcrunch.com/2025/11/06/meta-estimates-that-it-earns-10-of-its-revenue-from-scams-report-says/
It shouldn’t be controversial to say AI should always remain in human control – that we humans should remain at the top of the food chain. That means we need to start getting serious about guardrails, now, before superintelligence is too advanced for us to impose them.”” / X https://x.com/mustafasuleyman/status/1986834581576941763
It’s an understatement to say a lot of us haven’t felt good about our relationship with technology. So the absolute last thing we should be doing is making that relationship romantic. https://x.com/mustafasuleyman/status/1985389776330244108
Microsoft AI chief says only biological beings can be conscious https://www.cnbc.com/2025/11/02/microsoft-ai-chief-mustafa-suleyman-only-biological-beings-can-be-conscious.html
Microsoft forms superintelligence team under AI head Mustafa Suleyman https://www.cnbc.com/2025/11/06/microsoft-forms-superintelligence-team-under-ai-head-mustafa-suleyman-.html
Towards Humanist Superintelligence | Microsoft AI https://microsoft.ai/news/towards-humanist-superintelligence/
Microsoft’s $15.2 billion USD investment in the UAE – Microsoft On the Issues https://blogs.microsoft.com/on-the-issues/2025/11/03/microsofts-15-2-billion-usd-investment-in-the-uae/
Nvidia’s Jensen Huang: ‘China is going to win the AI race,’ FT reports | Reuters https://www.reuters.com/world/asia-pacific/nvidias-jensen-huang-says-china-will-win-ai-race-with-us-ft-reports-2025-11-05/
Nvidia’s Jensen Huang: ‘China is going to win the AI race,’ FT reports https://finance.yahoo.com/news/nvidias-jensen-huang-says-china-211900769.html
Strengthening ChatGPT’s responses in sensitive conversations | OpenAI https://openai.com/index/strengthening-chatgpt-responses-in-sensitive-conversations/
Built to benefit everyone | OpenAI https://openai.com/index/built-to-benefit-everyone/
Former OpenAI Exec Explains Why He Tried to Do a Coup Against Sam Altman https://gizmodo.com/former-openai-exec-explains-why-he-tried-to-do-a-coup-against-sam-altman-2000680769
OpenAI CFO Would Support Federal Backstop for Chip Investments https://www.wsj.com/video/openai-cfo-would-support-federal-backstop-for-chip-investments/4F6C864C-7332-448B-A9B4-66C321E60FE7
I would like to clarify a few things. First, the obvious one: we do not have or want government guarantees for OpenAI datacenters. We believe that governments should not pick winners or losers, and that taxpayers should not bail out companies that make bad business decisions or”” / X https://x.com/sama/status/1986514377470845007
I have been writing for years about the fact that we are not ready for the destruction of costly signalling mechanisms. Writing used to be a way of measuring effort, ability and diligence. We still have no easy substitute. https://x.com/emollick/status/1985854486317822204
Coca-Cola | Holidays are Coming, Fantastical :90 – YouTube https://www.youtube.com/watch?v=eoXX905YK6M
Today, we launched Helios, a technological marvel redefining the possible. Helios is the most accurate quantum computer in the world, with 98 of the highest fidelity physical qubits ever released, and 48 error-corrected logical qubits. Learn more: https://x.com/QuantinuumQC/status/1986172816241402189
Tesla shareholders approve $1 trillion pay package for Musk | CNN Business https://edition.cnn.com/2025/11/06/business/musk-trillion-dollar-pay-package-vote
Tesla Shareholders Approve Elon Musk’s $1 Trillion Pay Package – WSJ https://www.wsj.com/business/autos/elon-musk-tesla-pay-package-vote-9abd5a73?st=d8Surv&reflink=desktopwebshare_permalink&mod=tldr
For an AI assistant to be truly personal, it’s important to keep as much data as possible locally on the users device, and not on Perplexity servers. Account credentials, such as passwords and credit card information, are also stored locally on the user’s device.”” / X https://x.com/perplexity_ai/status/1985376891763925064
Perplexity launched Perplexity Patents, a new IP intelligence research Agent! “”While in beta, Perplexity Patents will be free for all users. Pro and Max subscribers will receive additional usage quotas and model configuration options.”” Now I want the same for News sources 👀 https://x.com/testingcatalog/status/1983885677835014270
Auth0 | Securing AI Agents | The New Identity Challenge | Auth0 https://auth0.com/resources/whitepapers/securing-ai-agents-the-new-identity-challenge
A Short Lesson in Simpler Prompts – nilenso blog https://blog.nilenso.com/blog/2025/11/04/a-short-lesson-in-simpler-prompts/
AI laziness remains one of the wildest common failure states when you think about it. Also one of the great examples of how much pseudo-humanity comes from training on the corpus of all human writing. The number of emails and chat messages exhibiting laziness beats post training”” / X https://x.com/emollick/status/1985507610963968447
Biggest gap between a brilliant passage written about a work of art and what you might expect the art to look like based on the passage? From Walter Benjamin (the painting in the reply) https://x.com/emollick/status/1984700905086722185
Finally, we found about 10% of tasks to have serious errors (though this isn’t an unusual rate). Some tasks have incorrect answer keys, some evaluation functions are too strict or too lax, and some instructions are fatally ambiguous.”” / X https://x.com/EpochAIResearch/status/1985441142343942242
The challenge in learning using AI is very similar to the same learning issue from search. When we are given answers we think we learn, but we don’t. Learning is work. However, things like the “learning modes” from the AI providers help, as does using AI for tutoring not answers”” / X https://x.com/emollick/status/1984254177246331238
The fact that no current AI models, often including GPT-5, believe in the existence of GPT-5 is amusing, frustrating & a very good example of why a lack of continuous learning hampers AI utility. They are often incredulous about other events that have happened this year as well https://x.com/emollick/status/1985822277485613496
Context engineering is the art of giving AI systems the right information at the right time. Most engineers focus on initial retrieval, that is, finding relevant documents from a large corpus. But retrieval is just the first step. The real challenge is prioritization. A reranker”” / X https://x.com/douwekiela/status/1985756688000163892
This is a surprisingly revealing test prompt: “Write a paragraph that startles me with its brilliance and really demonstrates your capabilities across as many dimensions as possible. Then explain what you did.” Claude excels at writing, GPT-5 Pro nails intellectual tricks, etc. https://x.com/emollick/status/1984827923363230182
AI Music’s Wild Ride: Udio’s Licensing Drama Unfolds https://www.therundown.ai/p/the-new-rules-of-ai-music
UNIVERSAL MUSIC GROUP AND UDIO ANNOUNCE UDIO’S FIRST STRATEGIC AGREEMENTS FOR NEW LICENSED AI MUSIC CREATION PLATFORM – UMG https://www.universalmusic.com/universal-music-group-and-udio-announce-udios-first-strategic-agreements-for-new-licensed-ai-music-creation-platform/
Xania Monet is the first AI-powered artist to debut on a Billboard airplay chart, but she likely won’t be the last | CNN https://edition.cnn.com/2025/11/01/entertainment/xania-monet-billboard-ai
I don’t think how people are tracking how quickly this is happening, for better or worse. https://x.com/emollick/status/1985132904771399899
Beloved Bodega Cat Reportedly Killed by Driverless Waymo https://futurism.com/advanced-transport/driverless-waymo-killed-cat
Remote Labor Index: Measuring AI Automation of Remote Work https://arxiv.org/pdf/2510.26787
🚀 new 🌤️ lighteval release and our biggest yet! • new benchmark finder to explore all available tasks • inspect-ai integration from @AISecurityInst → more stable and easier to add benchmarks • share your evals and insights with the community on the @huggingface hub • new https://x.com/nathanhabib1011/status/1985720151673880923
There will be no federal bailout for AI. The U.S. has at least 5 major frontier model companies. If one fails, others will take its place.”” / X https://x.com/DavidSacks/status/1986476840207122440
It is crazy that just ~2 years ago OAI board approached Dario to be CEO after firing Sam.
Thread by @ArfurRock on Thread Reader App – Thread Reader App https://threadreaderapp.com/thread/1984037162216825007.html
interesting post from @boazbaraktcs: https://x.com/sama/status/1985841697067319510
The big article on data centers in the New Yorker is pretty good, which I wasn’t expecting given the reaction on X. Lots of good and bad: and covering both bubble & non-bubble arguments. It also featured the best version of “I spoke to a local farmer about a data center” https://x.com/emollick/status/1985195665132040621
AI people are pretty good at renaming concepts to make them sound much more science fictional (also they just take a lot of names from science fiction). Emergence, grokking, alignment, transcendence, compute, latent space, oracles… now get ready to refer to power as electrons”” / X https://x.com/emollick/status/1986209965653303435
It would be useful if there was general agreement that the current range of AI capabilities will result in good & bad things, and we had more specific efforts (parallel to the larger existential discussions) to make the good stuff work for more people & mitigate the obvious bad.”” / X https://x.com/emollick/status/1984113533358063768
Taking Bold Steps to Keep Teen Users Safe on Character.AI https://blog.character.ai/u18-chat-announcement/
Of the many processes impacted by AI, innovation/design thinking seems like a key one in need of urgent change. Some aspects remain (building empathy), but many of the constraints change dramatically with AI (as our research shows). Careful thought needed to define a new process. https://x.com/emollick/status/1984653211936993723
Not just statements. When the Fed chair frowns during a press conference, the S&P 500 drops 0.53 basis points for the next 3 minutes. https://x.com/emollick/status/1983728197209616418
Australians have been promised three free hours of solar power a day. Here’s what you need to know | Energy | The Guardian https://www.theguardian.com/environment/2025/nov/04/australia-free-solar-power-scheme-how-when-houshold-bills
Google and Mombak collaborate on CO2 removal https://blog.google/outreach-initiatives/sustainability/mombak-co2-removal/
Image scraping is finally here! Firecrawl’s latest v2 endpoint now lets you scrape visual content from the web to build multimodal LLM apps, fine-tune LLMs, and more. You can also apply specific filters like resolutions, aspect ratios, or image types. 66k+ stars on GitHub! https://x.com/_avichawla/status/1985233254694416743
Comet was designed with privacy and security at its core. Today, we’re introducing new features that make it even easier to see and control privacy in Comet. https://x.com/perplexity_ai/status/1985376841021174184
Nvidia’s China plans hit roadblock as US moves to block sale of scaled-down AI chips to Beijing: Report | Today News https://www.livemint.com/news/us-news/nvidias-china-plans-hit-roadblock-as-us-moves-to-block-its-sale-of-scaled-down-ai-chips-to-beijing-report-11762483101906.html
The Department of Commerce has allowed Microsoft to ship NVIDIA GPU’s to the UAE for the first time. Brad Smith announced this today in Abu Dhabi. He said MS received the license in September, and will spend $7.9 billion on datacenters in the UAE over the next four years. https://x.com/AndrewCurran_/status/1985325278823125483
Sam’s clarification is good and important. Furthermore – I don’t think it can be overstated how critical compute will become as a national strategic asset. It is so important to build. It is vitally important to the interests of the US and democracy broadly to build tons of it”” / X https://x.com/jachiam0/status/1986583797492818244
All palaces are temporary palaces All theories are provisional theories”” / X https://x.com/sama/status/1984001525556097507
The US government should subsidize Open AI rather than OpenAI”” / X https://x.com/hardmaru/status/1986623547746492716
The government has played a role in critical infrastructure builds. Our public submission (posted on our blog) shares our thinking and suggests ideas for how the US government can support domestic supply chain/manufacturing. This is very in line with everything we have heard”” / X https://x.com/sama/status/1986917979343495650
Individual releases of open AI models only matter in the short term. These models becomes obsolete without continued releases (look at Llama versus newer Chinese models), because the capability/cost improvement curve is steep and you don’t want to use an older model forever. https://x.com/emollick/status/1984993332251263061
AI-Generated Billboard-Charting ‘Artist’ Xania Monet Creator Defends Music From Backlash https://www.forbes.com/sites/conormurray/2025/11/05/creator-behind-billboard-charting-ai-artist-xania-monet-defends-her-music-against-backlash-from-kehlani-and-more/
Baby Shoggoth Is Listening – The American Scholar https://theamericanscholar.org/baby-shoggoth-is-listening/
Common Crawl Is Doing the AI Industry’s Dirty Work – The Atlantic https://www.theatlantic.com/technology/2025/11/common-crawl-ai-training-data/684567/?gift=iWa_iB9lkw4UuiWbIbrWGQv84IP0_-K67yuVC013Fx4
Companies selling the dream of autonomous household humanoid robots today would be better off embracing reality and selling “remote operated household help”. Have teams of employees running them 24/7, with the option to reduce their workload as autonomous behaviors become viable.”” / X https://x.com/ID_AA_Carmack/status/1985390721315324380
The “AI will replace radiologists” prediction remains a rich example. A lot of folks have pointed out the problems with confusing a task (“reading a scan”) with a job (“radiologist”) with many tasks. That is true. But there was a human problem. Radiologists rejected (pre-LLM) AI”” / X https://x.com/emollick/status/1984696156140470530
LLMs memorize a lot of training data, but memorization is poorly understood. Where does it live inside models? How is it stored? How much is it involved in different tasks? @jack_merullo_ & @srihita_raju’s new paper examines all of these questions using loss curvature! (1/7) https://x.com/GoodfireAI/status/1986495330201051246
Pro tip to avoid public embarrassment: vibe test your models aggressively for things you can’t easily measure. Exhibit A: during SmolLM3, we accidentally nuked all the system messages in our training data and the model had no clue who it was 🙈 https://x.com/_lewtun/status/1985995034970214676
Leading AI Researcher Is Raising $1 Billion to Build AI Models With EQ – Business Insider https://www.businessinsider.com/researcher-raising-1-billion-to-build-ai-models-with-eq-2025-10
A tale in three acts: https://x.com/sama/status/1984023663642087831
So far, Grokipedia is very similar to its apparent source of Wikipedia, but articles tend to be longer with fewer references and more complex writing. https://x.com/emollick/status/1986081027551543315
Leave a Reply