Models, products and AI startups, near-duplicates collapsed to the most credible source
Stifel: Microsoft's AI bet is paying offStifel's Brad Reback says Azure growth is beating expectations and could stay above 50% near term, showing Microsoft's AI spending is paying off. He sees more risk at Oracle and says public AI lab results may reveal returns.
Attackers and defenders race on agentic AI securityA Bloomberg panel notes that both attackers and defenders are gaining AI-powered speed and scale, changing the threat landscape. Security leaders discuss emerging vulnerabilities and how to build resilience for increasingly autonomous enterprises.
Self-driving networks underpin the agentic enterpriseAs AI agents shift from insight to action, enterprises move toward more autonomous operations, per a Bloomberg session. Self-driving networks are framed as the foundation for this shift, improving resilience and keeping pace with machine speed.
Control may decide who captures agentic AI valueA Bloomberg discussion argues that as AI moves from generating answers to taking action, control over data, infrastructure and workloads may determine who captures the value of agentic AI.
MIT Professor Says AI Is Taking Over Your BrainAn opinion piece asks whether AI is taking over your brain and sparked discussion on Hacker News. The article body was not retrievable, only a comment link and page styling.
Only 40% of AI-cited statistics trace to a studyA test posed 200 everyday questions requiring numbers to ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews, then traced their citations. Only 40% of cited statistics could be traced back to an original study.
Startup &Truth applies formal verification to AI-generated codeStartup &Truth is joining the race to automate formal verification, which uses math to prove software works as intended. The approach is seen as a fix for the reliability problems of vibe coding, though the process is painstaking.
HN Essay Warns Against the 'LLM Voice' in WritingAn essay on Hacker News names the "LLM voice": readers often sense a text came from a large model and react with an uncanny-valley aversion. The author doubts detection tools and advises writers not to let LLMs draft prose, since even human editing struggles to remove the voice; outlines are acceptable, but the prose should be written by the author.
Suffolk unveils $50M 100MAG AI innovation facilityConstruction firm Suffolk unveiled 100MAG, a $50 million facility at its Boston headquarters campus that consolidates its investments in data, AI, robotics and quantum technology to address rising costs and labor shortages in construction.
Nasdaq 100 Falls as AI Trade Shows New CracksThe Nasdaq 100 fell as Alphabet's record 13% drop on an ad sales miss and Microsoft's plan to raise at least $25 billion from a bond sale revived doubts over returns on massive AI spending.
Redditor says local-AI setup cost him a jobA LocalLLaMA user says he was cut from a VC-backed AI company's hiring process after telling the co-founder he runs Qwen locally on a GPU. He acknowledges the 15-minute call may have failed for other reasons and offers it only as anecdata.
Johns Hopkins cancer mapping tech picked for Anthropic AI for ScienceJohns Hopkins' Alexander Szalay and Janis Taube were selected for Anthropic's AI for Science program to apply AstroPath, which combines astronomy, pathology and AI to map tumors. The program gives selected researchers free credits for Anthropic's Claude models; it began with biology and life sciences and has expanded to other fields.
AI agents accelerate into enterprise softwareBloomberg Intelligence examines the shift from generative AI tools to systems that can reason, decide and act, and what that means for enterprise software and technology investment.
Proton CEO's aggressive AI push sparks internal backlashReporting says Proton CEO Andy Yen is aggressively pushing AI internally: a feedback email to remove obstacles, software tracking employees' AI-generated code, and no engineer hiring this year. Staff will reportedly be re-interviewed on AI use, with implications for their jobs.
AI labs in progressives' sights if Democrats win HousePolitico reports that progressive lawmakers could target AI labs if Democrats win control of the House.
Study: embeddings plus simple classifiers beat Jev modelA Voyage AI employee reported that using an embedding model plus a simple sklearn classifier was cheaper and significantly more accurate than the Jev decision model when labeled data is available, cutting error by 37%. The author noted the gain comes from training labels that Jev does not use, so zero-shot cases may differ.
Michigan uses AI to rank inmate risk amid violenceMichigan's corrections department plans an AI system to rank its 32,000 inmates from security Level 1 to 5 in real time, replacing manual worksheets, amid at least four recent suspected cellmate killings. The corrections officers' union president argues the real problem is a shortage of Level 4-5 beds, not classification accuracy.
Mistral releases weights for trillion-parameter Mistral Large 4Mistral released the weights of its trillion-parameter Mistral Large 4, pitched at sovereign cyber defense. It claims performance comparable to GLM-5.3 while using 47% fewer tokens per task, and says ML4 tops other open models on Artificial Analysis' Coding Agent Index.
Anthropic files S-1 with valuation seen at $1.5T to $2TAnthropic has filed its S-1 to list on Nasdaq, possibly trading next month at a valuation of $1.5 trillion to $2 trillion. On Decoder, SemiAnalysis analyst Joey Brookhart breaks down what the filing reveals about frontier-lab economics: complex finances built on very ambitious assumptions about AI's future.
Mistral says Le Chonk leads open-weight raceFrench AI lab Mistral released a model called Le Chonk, claiming it matches or outperforms top Chinese open-weight models from DeepSeek and Z.ai in coding and cybersecurity. The release is framed as Europe leapfrogging China in the open-weight model race.
Thoughtworks CTO: Why Good Developers Are Wary of Agentic DevelopmentThoughtworks CTO Rachel Laycock writes that strong developers' caution about coding agents is not just love of coding or job fears: they know built is not the same as working in production. Their focus on building the right thing, and software that stays maintainable and reliable, explains the hesitancy.
Caterpillar shares plunge after earnings only modestly beat expectationsCaterpillar shares plunged after the company reported earnings that only modestly beat expectations. The drop adds to investor worries about mounting challenges to the AI infrastructure boom.
Op-ed Questions Trump Order Renaming AI as 'Super Intelligence'An op-ed in the Hillsdale Collegian says Trump signed an executive order last month replacing "artificial intelligence" with "super intelligence" in federal communications; the White House says the term better represents a promised future of US technological progress. The author argues the rename is no real qualitative change, implies machines surpassing humans, and warns about technology eroding freedom.
NYT: Chinese intelligence targeted Nvidia CEO Jensen HuangThe New York Times reports US intelligence believes Alice Huang of Singapore-based Megaspeed, close to Nvidia CEO Jensen Huang, passes information to Beijing; Megaspeed is also suspected of helping China obtain Nvidia chips. A Malaysian affiliate bought at least $4.8B of Nvidia-based AI hardware. Nvidia says it wasn't warned; Megaspeed opened an internal probe.
IBM ships Granite Switch 4.1.1 patch releaseIBM released Granite Switch 4.1.1, a backwards-compatible patch that needs no migration. It updates the Speech ASR adapter to handle 10 languages with native punctuation and capitalization, and improves the Decision adapter. Since the base model was untouched, only the adapters were retrained; an Adapter SDK is planned.
Analysis: Anthropic's S-1 justifies massive compute spendingThe Verge argues Anthropic's S-1 gives the company financial justification to spend as much as possible on compute. The piece says Anthropic is pinning its future on dramatic growth, since only that scale of growth can support its spending and avoid bankruptcy.
Cloudera launches Anywhere Cloud, Blueprints for hybrid AIAt its Evolve conference in New York, Cloudera unveiled Anywhere Cloud and Blueprints along with AI Agent Studio for building production-ready AI agents. It is betting on cloud, on-premises and hybrid deployment as its differentiator versus Databricks, Snowflake and hyperscalers, citing data sovereignty and compliance needs.
OpenAI's Astra becomes main target for Chinese AI labs after Anthropic tightens defensesAccording to The Decoder, Chinese AI labs are now mainly targeting OpenAI's most powerful model, GPT-6 Astra, rather than Anthropic's Claude, after Anthropic tightened its defenses.
Amazon adds 100+ shopping features to AlexaAhead of the holidays, Amazon added over 100 shopping capabilities to Alexa, including list and subscription management, editing delivery instructions, and answering Prime billing questions. Amazon says agentic-feature usage has grown 8x this year, and 400M+ consumers used Alexa for shopping in the past 12 months.
Deepgram acquires Daily to strengthen voice AI infrastructureVoice AI company Deepgram is acquiring Daily, the maker of the open-source Pipecat framework for voice and video pipelines. The deal amount was not disclosed, and both firms pledged Pipecat will remain open source and vendor-neutral. Daily had raised over $62M; both startups were in Y Combinator's 2016 batch.
The A.I. Daily