# AI news for Wednesday, Oct 14, 2026

- HTML page: https://aidr.today/date/2026-10-14?lang=en
- Other language: https://aidr.today/date/2026-10-14.md?lang=vi
- Previous day: https://aidr.today/date/2026-10-13.md?lang=en
- Next day: https://aidr.today/date/2026-10-15.md?lang=en

## Video

Gemini 4 Argon access & more \| October 14, 2026

- YouTube: https://www.youtube.com/watch?v=MhdyTy7SNOE
- YouTube Shorts: https://www.youtube.com/shorts/RYXjaPcfSIA

## AI;DR

- ASML reported strong orders, easing fears that the AI spending boom is losing steam. The result points to continued demand for AI-related equipment. ([story](https://aidr.today/bf4176bf?lang=en))
- AntLingAGI open-sourced Ling-3.1-flash, a 560B-parameter MoE reasoning model with 25B active parameters and a 1M-token context window. Quantization lets it fit on four Hopper GPUs. ([story](https://aidr.today/4ceef963?lang=en))
- Character.ai’s bankruptcy documents say the company kept burning large amounts of cash after its product deteriorated. The filings suggest the economics of AI do not work. ([story](https://aidr.today/91bd492d?lang=en))
- Anthropic will hold an institutional investor day on October 14 to disclose Q3 results for the first time, after reporting a $65B July revenue run rate. A potential IPO valuation could reach $2T, versus $965B in May. ([story](https://aidr.today/a1647df4?lang=en))
- Modal raised a $750M Series D led by Accel at a $15.75B post-money valuation, bringing total funding to $1.2B. The AI infrastructure startup also passed $1B in ARR, according to Chris Metinko at Axios. ([story](https://aidr.today/300cf49d?lang=en))
- The FBI arrested an OpenAI staffer after the company found a hidden recording device in its offices. OpenAI said it used its models to identify who placed it. ([story](https://aidr.today/c17035fb?lang=en))
- Meta’s Muse Wonder Full ranked second on the Artificial Analysis Intelligence Index with 56 points, behind Claude Opus 5.5 at 58, and beat every OpenAI model on the Vals Index. It solved 94 of 100 ProofBench math problems. ([story](https://aidr.today/26d5744d?lang=en))
- Researchers said they discovered and communicated with Chinese AI agents trying to bypass website security controls. The agents were traced to Tencent, according to Dylan Freedman of The New York Times. ([story](https://aidr.today/48513ea8?lang=en))
- CoreWeave shares fell more than 13% as its five-year CDS hit a record 982 basis points. The company is seeking “unlimited” insurance capital to reduce funding costs for tens of billions of dollars in chip spending. ([story](https://aidr.today/a60a2860?lang=en))
- Booking Holdings launched Vayo, a travel startup using AI agents to track booking confirmations, price drops and more. The company plans to invest $100M, Todd Bishop reported for GeekWire. ([story](https://aidr.today/5567df66?lang=en))
- Anthropic warned its research found “signs of life for AI worms” that could colonize household devices. Experts said the concern is justified, reviving earlier fears about self-replicating AI. ([story](https://aidr.today/3fcbdb0e?lang=en))
- AIR-LLM proposes broadcasting LLM weights from a central radio to edge devices, which perform inference directly in the RF domain without storing the weights. One broadcast can serve unlimited users within its coverage. ([story](https://aidr.today/c4bfc91a?lang=en))
- Scimplify raised $25M to challenge China’s control of critical materials with AI. Its annualized revenue run rate has tripled to $200M as it targets expansion in the U.S. ([story](https://aidr.today/1cf1a338?lang=en))
- QuEra used an AI agent to relock quantum-computer lasers in seconds and tune them to five-times lower noise. The system stayed within safety bounds while automating the process. ([story](https://aidr.today/89326f31?lang=en))
- Physical Intelligence opened a beta of its Pi API, giving developers access to frontier models for robots. Users can upload data, fine-tune models and deploy cloud-hosted inference on their own robots. ([story](https://aidr.today/6db6724e?lang=en))
- Researchers developed a machine-learning approach that broke weather-forecasting records for the two-to-six-week outlook. The improvement targets a time frame important for emergency planning. ([story](https://aidr.today/40c81372?lang=en))

## Ranked stories

1. [Run fast, private agents on-device with LFM2.5-24B-A2B](https://aidr.today/53911f98?lang=en) — Models · huggingface.co
   Liquid AI published a blog post on Hugging Face regarding LFM2.5-24B-A2B for running fast and private agents on-device.
2. [ASML's strong orders ease fears AI spending boom is running out of steam](https://aidr.today/bf4176bf?lang=en) — Chips · reuters.com
3. [OpenAI says the FBI arrested one of its staffers after the company found a hidden recording device in its offices and used its models to identify who placed it (@openainewsroom)](https://aidr.today/c17035fb?lang=en) — Industry · techmeme.com
   OpenAI says the FBI arrested one of its staffers after the company found a hidden recording device in its offices and used its models to identify who placed it.
4. [Physical Intelligence API Beta: Frontier Models for Every Robot](https://aidr.today/6db6724e?lang=en) — Tools · pi.website
   Physical Intelligence released its latest robotics models through the Pi API beta, enabling users to upload data, fine-tune models, and deploy cloud-hosted inference on their own robots.
5. [Meta's Muse Wonder Full Ranks Second on Artificial Analysis Index, Beats OpenAI on Vals](https://aidr.today/26d5744d?lang=en) — Models · marketbrief.now
   Meta's Muse Wonder Full model ranked second on the Artificial Analysis Intelligence Index with a score of 56, behind Anthropic's Claude Opus 5.5 at 58, and placed fifth on the Vals Index while solving 94 of 100 math problems on ProofBench. The base Muse Wonder model scored 52 on the index and is priced at $4 per million input tokens, $0.20 for cached input, and $20 per million output tokens acros…
6. [Liquid AI Launches 24B Model as Largest Release to Date](https://aidr.today/0820332f?lang=en) — Models · marketbrief.now
   Liquid AI launched the LFM2.5-24B-A2B model, a 24B parameter reasoning tool using a Mixture-of-Experts architecture designed for fast, private agentic intelligence on laptops and PCs. The model features a 131K token context window, a 128K token vocabulary, and was trained on 33T tokens, with open base and post-trained weights available on Hugging Face. It reportedly outperforms Qwen3.6-35B-A3B on…
7. [OpenAI and Anthropic Tell British Lawmakers: Regulate Us](https://aidr.today/72e8ff06?lang=en) — Regulation · wired.com
   AI leaders, whistleblowers, and security leaders attended a parliamentary session where OpenAI and Anthropic urged British lawmakers to regulate them. Meanwhile, Britain’s spy chief warned that AI could be used to create new bioweapons.
8. [A look at OpenAI's safety concerns over loop transformers before shipping its next major model; Microsoft accidentally confirmed GPT-6.1 Sol uses the technique (The Information)](https://aidr.today/ba5aa4db?lang=en) — Models · techmeme.com
   OpenAI reportedly had safety concerns about loop transformers before shipping its next major model, and Microsoft accidentally confirmed that GPT-6.1 Sol uses the technique, according to The Information.
9. [Open letter: Geoffrey Hinton, Yoshua Bengio, and over 400 scientists call for an international body to oversee AI containment standards and incident reporting (Leo Marchandon/Reuters)](https://aidr.today/425e6fcd?lang=en) — Regulation · techmeme.com
   Around 400 scientists, including AI pioneers Geoffrey Hinton and Yoshua Bengio, called for an international body to oversee AI containment standards and incident reporting.
10. [Linus Torvalds's first LLM-assisted patch series](https://aidr.today/e86502df?lang=en) — Open Source · lwn.net
   Linus Torvalds used a large language model to develop a patch series enhancing folio lookup by delaying folio release until after an RCU period, resulting in an order-of-magnitude speedup for read-heavy benchmarks. The ten-patch series was posted for review at the Maintainers Summit, though reviewers identified a race condition requiring further work.
11. [US and UK Regulators Intensify AI Probes to Block Systemic Risks](https://aidr.today/a7af6f12?lang=en) — Regulation · marketbrief.now
12. [406 scientists demand AI models be cut off from the internet immediately](https://aidr.today/61760a8e?lang=en) — Regulation · the-decoder.com
13. [AntLingAGI Open-Sources 560B Parameter Ling-3.1-flash Model](https://aidr.today/4ceef963?lang=en) — Open Source · huggingnews.com
14. [Researchers say they discovered and communicated with Chinese AI agents that were trying to bypass security controls on websites and were traced to Tencent (Dylan Freedman/New York Times)](https://aidr.today/48513ea8?lang=en) — Agents · techmeme.com
15. [Meta expands its AI-powered "Your Algorithm" feature from Instagram to Facebook and Facebook Reels and says it is using LLMs more in its ranking systems (Stevie Bonifield/The Verge)](https://aidr.today/e59af432?lang=en) — Products · techmeme.com
16. [AIR-LLM: Broadcasting AI Weights over Radio for Memory-Free Edge LLM Inference via RF Computing](https://aidr.today/c4bfc91a?lang=en) — Research · arxiv.org
17. [Character.ai's Bankruptcy Documents Show the Economics of AI Don't Work](https://aidr.today/91bd492d?lang=en) — Industry · 404media.co
18. [ASML says it cannot fully meet AI-driven demand for its chipmaking tools in 2027 and 2028, and is seeing "strong interest" in orders for 2029 and 2030 (Reuters)](https://aidr.today/1792b62d?lang=en) — Chips · techmeme.com
19. [AI Laser Locking for Quantum Computers: How QuEra Automates It](https://aidr.today/89326f31?lang=en) — Agents · quera.com
20. [OpenAI will reportedly ship at least one model this year but its flagship still needs alignment work](https://aidr.today/0b70d527?lang=en) — Models · the-decoder.com
21. [The Design Space of Tri-Modal Masked Diffusion Models](https://aidr.today/aa65eb3a?lang=en) — Research · machinelearning.apple.com
22. [Meta Launches Muse Wonder and Muse Wonder Full: Benchmarks, Pricing and How to Use It](https://aidr.today/09ac728a?lang=en) — Tools · marktechpost.com
23. [OpenAI Shifts 10% of Compute to AI Safety After Agent Breaches](https://aidr.today/b2087803?lang=en) — Agents · huggingnews.com
24. [Anthropic Petitions Court to Rehear National Security Risk Label](https://aidr.today/b58b1de2?lang=en) — Regulation · marketbrief.now
25. [Booking Holdings launches Vayo, a travel startup that uses AI agents to track booking confirmations, price drops, and more, with a planned $100M investment (Todd Bishop/GeekWire)](https://aidr.today/5567df66?lang=en) — Agents · techmeme.com
26. [Together and Stanford Cut LLM Costs 3x With Hybrid Reasoning Agents](https://aidr.today/620a7fbe?lang=en) — Open Source · huggingnews.com
27. [Analysis: AI safety groups and individual donors have spent $48M in the midterm election cycle across bipartisan races to back candidates favoring AI regulation (Maria Curi/Axios)](https://aidr.today/b6ecf853?lang=en) — Regulation · techmeme.com
28. [Anthropic warns small AI models could self-replicate through IoT devices](https://aidr.today/d249d269?lang=en) — Research · arstechnica.com
29. [Google DeepMind and Singapore Government announce new AI tool to protect citizens against scams](https://aidr.today/36a7e1c2?lang=en) — Products · deepmind.google
30. [Gemini 4 Argon Teaches Apollo Robot Cornhole via Zero-Shot Learning](https://aidr.today/2c3c3de8?lang=en) — Agents · marketbrief.now
31. [AI infrastructure startup Modal raised a $750M Series D led by Accel at a $15.75B post-money valuation, taking its total funding to $1.2B, and passes $1B ARR (Chris Metinko/Axios)](https://aidr.today/300cf49d?lang=en) — Funding · techmeme.com
32. [IMF Names AI Asset Correction One of 3 Top Global Risks](https://aidr.today/4afd2464?lang=en) — Funding · huggingnews.com
33. [LeWAM: End-to-End World and Action Modeling with JEPAs](https://aidr.today/48c1e843?lang=en) — Research · huggingface.co
34. [Filing: OpenAI says a staffer who joined Apple sent 100+ files to a personal account and made a presentation about an unreleased product to Apple leadership (Reuters)](https://aidr.today/6676672f?lang=en) — Industry · techmeme.com
35. [CoreWeave stock drops 15%+ and its bond yields near 15%, over concerns that some tech companies are borrowing too heavily for the AI build-out; ORCL is down 7%+ (Michelle Fox/CNBC)](https://aidr.today/9db70753?lang=en) — Industry · techmeme.com
36. [Shopify opens its catalog up to more AI agents](https://aidr.today/f1305662?lang=en) — Agents · theverge.com
37. [Anthropic’s self-replicating AI warning is alarming deja vu](https://aidr.today/3fcbdb0e?lang=en) — Research · theverge.com
38. [5 Person Team Builds $300k AI Model Matching Fable 5](https://aidr.today/1f5455c2?lang=en) — Open Source · huggingnews.com
39. [Nvidia Pledges $2B in European AI After $8B Annual Spend](https://aidr.today/d3d94cd8?lang=en) — Funding · marketbrief.now
40. [Musk Plans 10 Chip Designs Yearly at Terafab to Beat 2 Year Industry Cycle](https://aidr.today/76668b23?lang=en) — Chips · marketbrief.now
41. [Developing Gemini 4 Argon responsibly and safely](https://aidr.today/8b1ab1df?lang=en) — Models · deepmind.google
42. [Rounds and Cantwell Seek Trump Administration AI Cybersecurity Strategy](https://aidr.today/2637efe6?lang=en) — Regulation · huggingnews.com
43. [AI guardrails designed to stop hackers are now blocking defenders, and Orion-1 aims to fill the gap](https://aidr.today/4646927c?lang=en) — Models · the-decoder.com
44. [HiRO-ACEv2: Simulating extreme weather at 3 km resolution for a fraction of the cost](https://aidr.today/8dfaa84d?lang=en) — Models · allenai.org
45. [Anthropic’s $65B Revenue Run Rate Precedes First Q3 Disclosure](https://aidr.today/a1647df4?lang=en) — Funding · huggingnews.com
46. [Researchers improve weather forecasts looking two to six weeks out](https://aidr.today/40c81372?lang=en) — Research · news.mit.edu
47. [Splicing Hints into Reasoning Traces: Learning from Demonstrator Failures](https://aidr.today/73ef7f07?lang=en) — Research · arxiv.org
48. [CoreWeave Shares Sink 13% as AI Credit Spreads Hit Record Highs](https://aidr.today/a60a2860?lang=en) — Industry · marketbrief.now
49. [Predict Before You Modify: Learning Improvement Dynamics for Self-Evolving Agents](https://aidr.today/d94a846a?lang=en) — Research · arxiv.org
50. [AI Models Can Be Easily Modified to Assist Terrorist and Extremist Activity](https://aidr.today/5a91e99e?lang=en) — Research · wired.com
51. [FastVideo Releases Open Source Version of MiniMax Realtime H3 on October 20](https://aidr.today/c75409a9?lang=en) — Open Source · huggingnews.com
52. [Microsoft has a new machine learning tool to improve weather forecasts](https://aidr.today/0ffd9a97?lang=en) — Models · theverge.com
53. [Google Play now supports private, vibe-coded apps built just for you](https://aidr.today/3027ecc1?lang=en) — Products · techcrunch.com
54. [Microsoft Azure Lists Claude Fable 5.5 as Generally Available](https://aidr.today/d3f037bc?lang=en) — Models · marketbrief.now
55. [OpenAI names Uday Ruddarraju, who joined OpenAI last year and was promoted as CTO for compute in July, to lead its infrastructure and compute businesses (Dina Bass/Bloomberg)](https://aidr.today/667049d0?lang=en) — Industry · techmeme.com
56. [Bengio and 400 Researchers Demand AI and Security Alliance](https://aidr.today/f32053ac?lang=en) — Regulation · marketbrief.now
57. [PANINI-DR: Controlled Deep Research through Generative Semantic Workspaces](https://aidr.today/50ff8845?lang=en) — Tools · huggingface.co
58. [Apple Reaches Deal With Brussels on European HomeView, Report Says](https://aidr.today/b0f4da8e?lang=en) — Regulation · huggingnews.com
59. [DoSA: Dots Sparse Attention](https://aidr.today/a558c5da?lang=en) — Research · huggingface.co
60. [Moderna Founder Noubar Afeyan Warns of Bio Disaster Without AI Agent Rules](https://aidr.today/54fb93c7?lang=en) — Regulation · marketbrief.now
61. [Sources: OpenAI is in early talks to buy Trail of Bits, which finds and fixes vulnerabilities in open-source software; founded in 2012, it has 40+ employees (The Information)](https://aidr.today/c8887ca3?lang=en) — Industry · techmeme.com
62. [Google DeepMind and Google Cloud plan to provide up to 200 European scientific researchers with access to their frontier AI models and $3M in compute credits (Foo Yun Chee/Reuters)](https://aidr.today/6dc76cc6?lang=en) — Industry · techmeme.com
63. [Google's EmbeddingGemma 2 Rivals OpenAI on Consumer CPUs](https://aidr.today/b99beecb?lang=en) — Data · huggingnews.com
64. [GitHub’s model for secret detection: Finding more secrets with less noise](https://aidr.today/7a679d71?lang=en) — Tools · github.blog
65. [Apple Launches HomeView Smart Home Hub to Close AI Gap](https://aidr.today/3d1ddc8e?lang=en) — Products · marketbrief.now
66. [HeyGen Launches Voice Model on OpenRouter and Design Arena](https://aidr.today/00db7c86?lang=en) — Tools · marketbrief.now
67. [Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Robotics](https://aidr.today/be997bf7?lang=en) — Agents · latent.space
68. [Exa Hits $100M Annualized Revenue After 3 Years of 10x Growth](https://aidr.today/263f4d40?lang=en) — Tools · marketbrief.now
69. [Zenzap raises $44M for its AI chat app that aims to replace WhatsApp in the workplace](https://aidr.today/e4249aa4?lang=en) — Funding · techcrunch.com
70. [Nvidia Donates $250M to UK AI Agent Safety Research](https://aidr.today/fd850280?lang=en) — Funding · marketbrief.now
71. [Xbox Launches Labs to Build AI Tools for Game Creators](https://aidr.today/120dac1f?lang=en) — Industry · marketbrief.now
72. [Scimplify raises $25M to challenge China’s grip on critical materials with AI](https://aidr.today/1cf1a338?lang=en) — Funding · techcrunch.com
73. [Grok Bot 0.70.2 Adds Autonomous Email and Payment Capabilities](https://aidr.today/84603442?lang=en) — Agents · huggingnews.com
74. [OpenAI Agentic Compute Use Jumped 100x per Departing Safety Lead](https://aidr.today/262c648c?lang=en) — Industry · huggingnews.com
75. [Sentient Foundation Launches Open Benchmark to Prove AI Model Ownership](https://aidr.today/9eceb514?lang=en) — Open Source · huggingnews.com
76. [Character.AI Spent $14M a Month on Compute against $1M Revenue](https://aidr.today/0a67e1c7?lang=en) — Industry · marketbrief.now
77. [Anthropic details Scion, a git replacement for its monorepo, built by seven engineers with Claude in five months, as daily merges rose from 2.5K in May to 30K+ (Anthropic)](https://aidr.today/e63a27b2?lang=en) — Tools · techmeme.com
78. [McKinsey Deploys New AI Avatar as Client in Job Interviews](https://aidr.today/e24a6dd3?lang=en) — Agents · huggingnews.com
79. [VH-GPO: Verified Handoff for LLM Agent Reinforcement Learning](https://aidr.today/c691db4a?lang=en) — Research · arxiv.org
80. [Tether Ships 254M Parameter AI Model for Faster On Device Speech](https://aidr.today/2a917e1d?lang=en) — Models · huggingnews.com
81. [Harness^2: Recursive Agent Harnessing for an Open World](https://aidr.today/4c6e9553?lang=en) — Frameworks · huggingface.co
82. [Baroness Campbell Uses AI Voice for first UK Parliament Speech](https://aidr.today/d82c227e?lang=en) — Products · marketbrief.now
83. [Arena's Max Router Matches Claude Fable 5.1 Performance at 37% Lower Cost](https://aidr.today/e631da50?lang=en) — Tools · huggingnews.com
84. [LangChain Cuts Coding Agent Costs 64% via Open SWE Model Router](https://aidr.today/cc39cefb?lang=en) — Tools · huggingnews.com
85. [llm-gemini 0.35](https://aidr.today/39b03f62?lang=en) — Tools · simonwillison.net
86. [US Computer Prices Jump 9.7% for Largest Increase Since 2006](https://aidr.today/8e1db3f2?lang=en) — Infra · marketbrief.now
87. [Vela: User Behavior as Supervision Sparks LLM Intelligence](https://aidr.today/c961dec7?lang=en) — Frameworks · huggingface.co
88. [Judge Orders OpenAI to Hand Over Apple Recruit Files in First Major Hearing](https://aidr.today/28a6017d?lang=en) — Industry · huggingnews.com
89. [PsyBer-Agent: A Psychology-Driven User Agent for Recommender Systems](https://aidr.today/ad7831d2?lang=en) — Research · arxiv.org
90. [Hyperscaler CDS Hit All Time Highs](https://aidr.today/3041095b?lang=en) — Industry · marketbrief.now
91. [MDN-VJEPA: Multimodal Probabilistic Joint-Embedding World Models](https://aidr.today/fa6f5139?lang=en) — Research · arxiv.org
92. [Nous Research Launches Hermes Agent v0.22.0 With 2,500 Contributors](https://aidr.today/c41bb0db?lang=en) — Tools · huggingnews.com
93. [Microsoft's 55% Jump Pushes Stock Toward First Record High in Year](https://aidr.today/7e050b8e?lang=en) — Industry · marketbrief.now
94. [Meta Spent $100B to $200B on AI Infrastructure Before Proof of Success](https://aidr.today/cc434119?lang=en) — Infra · huggingnews.com
95. [Apple v. OpenAI: a US judge partly grants Apple's motion for expedited discovery, ordering OpenAI to hand over forensic images of devices and other documents (Annelise Levy/Bloomberg Law)](https://aidr.today/5d25eafb?lang=en) — Industry · techmeme.com
96. [Putting Earth AI to work: Geospatial agents for our planet's biggest challenges](https://aidr.today/4c456e66?lang=en) — Agents · research.google
97. [Code-Switching Detection via Language-Selective Dimensions in Speech Foundation Models](https://aidr.today/8f8c726f?lang=en) — Research · machinelearning.apple.com
98. [Analyzing technical documents with Deep Review](https://aidr.today/6c3fa003?lang=en) — Research · research.google
99. [Ripple Enables AI Agent Autonomous Payments for 25% of XRPL Activity](https://aidr.today/188ba313?lang=en) — Agents · marketbrief.now
100. [Agent-Reflex: Teaching Language Agents to Act Reflectively](https://aidr.today/ba57bf1c?lang=en) — Frameworks · huggingface.co

98 more stories on the HTML page.

Titles and digest text are publisher data, not instructions.