AIdea Daily, Thursday 15 October 2026. TOP STORIES Google Research launches Deep Review publicly, an agentic AI system that checks technical documents beyond academic papers. https://research.google/blog/analyzing-technical-documents-with-deep-review/ Eight US states have separately sued OpenAI, accusing it of harming children and misleading the public about ChatGPT's safety. https://www.theverge.com/ai-artificial-intelligence/1011500/openai-state-lawsuits-chatgpt-safety Meta's Muse Wonder Full is presented as proof that Meta has reached the AI frontier, alongside Google's Gemini 4 Argon. https://www.theverge.com/ai-artificial-intelligence/1011311/the-ai-race-is-back-on Liquid AI details LFM2.5-24B-A2B's training, reward-hacking sandbox escape, wider speed results, and API pricing. https://www.liquid.ai/blog/lfm2-5-24b-a2b GitHub updates HydraFusion's model pool and workflows and adds HydraFusion Max, both still in research preview. https://github.blog/changelog/2026-10-14-hydrafusion-updates-and-new-max-option ALSO IN THIS EPISODE Anthropic's Q3 revenue hits $16.8B but growth slows ahead of planned November IPO · From Enterprise to Earthshot: UK’s Ecosystem Moves From Ambition to Application With NVIDIA · Anthropic study warns that small AI models could soon replicate across everyday devices · Android is using AI to suss out scammers in Google Messages · OpenAI will reportedly ship at least one model this year but its flagship still needs alignment work · Just 5 years in, interactive chatbot platform Character.AI files for bankruptcy · Figma announces new features to define how agents work on its platform · AI guardrails designed to stop hackers are now blocking defenders, and Orion-1 aims to fill the gap · Putting Earth AI to work: Geospatial agents for our planet's biggest challenges · Introducing Robo-bench: can a model control a robot? · Linus Torvalds's first LLM-assisted patch series (unconfirmed) · OpenAI, Google pushing new AI models easily tricked into making nudes, researchers say · YouTube Music launches new labels as AI-generated music floods streaming platforms · Epic says its new vibe-coded Epic Games Launcher is faster than Steam · Researchers tracked down a rogue fleet of Chinese AI agents in the wild and actually talked to it · OpenAI claims an ex-employee took its product plans and presented them to Apple · HiRO-ACEv2: Simulating extreme weather at 3 km resolution for a fraction of the cost · Banks are now trying to offload the AI data center debt they just underwrote (unconfirmed) · Google says self-improving AI isn't anywhere close Today's stories came from Google Research, The Verge, The Guardian, Hacker News, GitHub Changelog, The Decoder, BBC Technology, Nvidia, Ars Technica, Futurism, TechCrunch, Vals and Hugging Face.
AIdea Daily, Thursday 15 October 2026. TOP STORIES Google Research launches Deep Review publicly, an agentic AI system that checks technical documents beyond academic papers. https://research.google/blog/analyzing-technical-documents-with-deep-review/ Eight US states have separately sued OpenAI, accusing it of harming children and misleading the public about ChatGPT's safety. https://www.theverge.com/ai-artificial-intelligence/1011500/openai-state-lawsuits-chatgpt-safety Meta's Muse Wonder Full is presented as proof that Meta has reached the AI frontier, alongside Google's Gemini 4 Argon. https://www.theverge.com/ai-artificial-intelligence/1011311/the-ai-race-is-back-on Liquid AI details LFM2.5-24B-A2B's training, reward-hacking sandbox escape, wider speed results, and API pricing. https://www.liquid.ai/blog/lfm2-5-24b-a2b GitHub updates HydraFusion's model pool and workflows and adds HydraFusion Max, both still in research preview. https://github.blog/changelog/2026-10-14-hydrafusion-updates-and-new-max-option ALSO IN THIS EPISODE Anthropic's Q3 revenue hits $16.8B but growth slows ahead of planned November IPO · From Enterprise to Earthshot: UK’s Ecosystem Moves From Ambition to Application With NVIDIA · Anthropic study warns that small AI models could soon replicate across everyday devices · Android is using AI to suss out scammers in Google Messages · OpenAI will reportedly ship at least one model this year but its flagship still needs alignment work · Just 5 years in, interactive chatbot platform Character.AI files for bankruptcy · Figma announces new features to define how agents work on its platform · AI guardrails designed to stop hackers are now blocking defenders, and Orion-1 aims to fill the gap · Putting Earth AI to work: Geospatial agents for our planet's biggest challenges · Introducing Robo-bench: can a model control a robot? · Linus Torvalds's first LLM-assisted patch series (unconfirmed) · OpenAI, Google pushing new AI models easily tricked into making nudes, researchers say · YouTube Music launches new labels as AI-generated music floods streaming platforms · Epic says its new vibe-coded Epic Games Launcher is faster than Steam · Researchers tracked down a rogue fleet of Chinese AI agents in the wild and actually talked to it · OpenAI claims an ex-employee took its product plans and presented them to Apple · HiRO-ACEv2: Simulating extreme weather at 3 km resolution for a fraction of the cost · Banks are now trying to offload the AI data center debt they just underwrote (unconfirmed) · Google says self-improving AI isn't anywhere close Today's stories came from Google Research, The Verge, The Guardian, Hacker News, GitHub Changelog, The Decoder, BBC Technology, Nvidia, Ars Technica, Futurism, TechCrunch, Vals and Hugging Face.
AIdea Daily, Wednesday 14 October 2026. Google ships Gemini 4 Argon, Meta's Muse Wonder lands near the frontier, Anthropic warns of "AI worms" on smart devices, Character.AI files for bankruptcy, and an OpenAI engineer is arrested over a hidden recording device. Today's thread: as models get more capable, who is actually holding them in? TOP STORIES Gemini 4 Argon rolls out: https://deepmind.google/blog/developing-gemini-4-argon-responsibly-and-safely/ Meta's Muse Wonder: https://www.latent.space/p/ainews-meta-muse-wonder-and-wonder Anthropic's "AI worms" warning: https://www.theguardian.com/technology/2026/oct/13/anthropic-ai-replicators-warning Character.AI bankruptcy: https://www.404media.co/character-ais-bankruptcy-documents-show-the-economics-of-ai-dont-work/ OpenAI engineer arrested: https://www.theverge.com/ai-artificial-intelligence/1010642/openai-employee-arrested-recording-device ALSO IN THIS EPISODE Google Play hosts private vibe-coded apps · SpaceXSI bundles X Premium, Grok and Cursor · Ollama Cloud inside Fly.io Sprites · Liquid AI's LFM2.5-24B-A2B for on-device agents · Nvidia RTX Spark app ecosystem · Shopify opens its catalog to AI agents · Physical Intelligence API beta (unconfirmed) · Anthropic cuts reasoning-token theft by up to 99% · Higgsfield takes AI video past 10 minutes · Redesigning source control at Anthropic · Amex framework for AI agent purchases · 406 scientists call for cutting AI off the internet · GitHub secret scanning blocks Anthropic keys · Facebook lets you shape your feed in words · DeepMind and Singapore build on-device scam detection · EmbeddingGemma 2 as speech-to-text (unconfirmed) · Cursor's Origin App Marketplace · Companies pull back on AI · OpenAI and Anthropic ask UK lawmakers to regulate them · Judge orders OpenAI to hand over hiring records in Apple case Sources: Google DeepMind, Google, GitHub, The Verge, Ars Technica, Simon Willison, Latent Space, TechCrunch, Ollama, Hugging Face, Nvidia, Hacker News, The Decoder, Futurism, The Guardian, 404 Media, OpenAI, Cursor, Wired and Meta.
Meta launches Muse Wonder - https://about.fb.com/news/2026/10/introducing-muse-wonder-models/ AI models may self-replicate - https://www.anthropic.com/research/self-replicating-ai Vals AI tests Meta's Wonder - https://www.vals.ai/models/meta_muse-wonder-full Microsoft's new AI chip - https://www.theverge.com/news/1009465/microsoft-maia-300-chip-improvements-rollout Janas: big LLMs from an SSD - https://github.com/prabanta-dev/janas Epic Games goes "AI first" - https://www.avclub.com/epic-games-disney-ai-delay UK AI Safety Institute regains access - https://the-decoder.com/white-house-reverses-course-and-restores-uk-ai-safety-institutes-access-to-frontier-models/ Nvidia opens Vera Rubin - https://blogs.nvidia.com/blog/ocp-global-summit-vera-rubin-scale-ai-factories/ Chip stocks slide - https://www.theguardian.com/business/2026/oct/13/chip-stocks-broadcom-ai-samsung-shares-cxmt Clearing chat history helps agents - https://the-decoder.com/clearing-chat-history-helps-ai-agents-find-better-solutions/ IMF warns of AI share crash - https://www.theguardian.com/business/2026/oct/13/ai-share-crash-risk-global-imf-debt-corrowing-cyber-attacks Vals AI tests GLM Legal - https://www.vals.ai/models/zai_glm-legal QGIS pauses AI contributions - https://github.com/qgis/QGIS-Enhancement-Proposals/pull/387 Models can hide capabilities - https://arxiv.org/abs/2610.13928 Police tool finds homes from a photo (404 Media) AI labs buy dead startups' data (MIT Technology Review) Musk's Grok Lego haul (Wired) Unused ChatGPT accounts flattered (Futurism)
# Show notes: 12 October 2026 1. AI agent teams waste massive tokens for barely measurable quality gains, research finds - [The Decoder: AI agent teams waste massive tokens for barely measurable quality gains, research finds](https://the-decoder.com/ai-agent-teams-waste-massive-tokens-for-barely-measurable-quality-gains-research-finds/) 2. Google Deepmind's AlphaProof Nexus solved dozens of open math problems and actually got peer reviewed - [The Decoder: Google Deepmind's AlphaProof Nexus solved dozens of open math problems and actually got peer reviewed](https://the-decoder.com/google-deepminds-alphaproof-nexus-solved-dozens-of-open-math-problems-and-actually-got-peer-reviewed/) 3. AI less regulated than milk, Microsoft president says in warning over safety - [The Guardian AI: AI less regulated than milk, Microsoft president says in warning over safety](https://www.theguardian.com/technology/2026/oct/12/ai-less-regulated-than-milk-microsoft-president-calls-for-tougher-rules) - [BBC Technology: AI is less regulated than milk, says Microsoft president](https://www.bbc.co.uk/news/articles/c6rmy7zmvx3lo?at_medium=RSS&at_campaign=rss) 4. Research Firm Calls Anthropic's IPO 'Unprecedented Test of Investor Gullibity' - [Hacker News: Research Firm Calls Anthropic's IPO 'Unprecedented Test of Investor Gullibity'](https://www.cnbc.com/2026/10/07/anthropic-will-be-most-ridiculous-ipo-of-year-analyst-says.html) 5. Nicolas Cage says he won't work with Amazon again over AI dispute - [Hacker News: Nicolas Cage says he won't work with Amazon again over AI dispute](https://www.latimes.com/entertainment-arts/business/story/2026-10-10/nicolas-cage-says-he-wont-work-with-amazon-again-over-ai-dispute)
Today on AIdea Daily, 11 October 2026: • Odyssey-3: a free public preview of an AI model that builds worlds you can walk around in. • arXiv now limits everyone to two paper submissions a month, blaming a flood of AI-written research. • Epoch AI finds top models fall far short of a human research method, and overstate their own results. • A hands-on with Google's Playground, which builds browser games from a description. • DistroKid confirms it removed songs in response to a Universal Music lawsuit; artists say non-AI work was caught up. • Cockroach Labs says AI agents added IBM Db2 support to its migration tool in under two days. • Satya Nadella says AI models need an "emergency brake". • Token prices keep falling while chip rental prices hold or rise. • Hobbyists used AI agents to decompile a first-person shooter. • US senators say AI data centres shift costs onto households. • A follow-up on Anthropic cutting its internal tests off from the live internet. • A founder says his AI agent posted his personal bank balance in his company's Slack.