Models, products and AI startups, near-duplicates collapsed to the most credible source
US employers face religious exemptions from workplace AIAs firms adopt generative AI, some employees are seeking exemptions from AI tools on sincerely held moral or religious grounds, though no such cases have reached court yet. Religious leaders and secular humanist groups are also urging attention to AI's effects on human agency.
User praises OrcaCyber-Nano-1.0 fine-tuneA LocalLLaMA user praised the Orcarouter team's OrcaCyber-Nano-1.0 fine-tune, saying it almost always one-shots bash command chains, answers faster than its base model, and built an internal chat app backend within a 120K context window. The user ran the 15.7GB version on 24GB of VRAM across two GPUs, noting it was used for general app development, not security work.
Z.ai's Jie Tang reveals GLM-5.5 training pipelineAccording to a social media post, Z.ai chief scientist Jie Tang revealed the training pipeline behind the upcoming GLM-5.5, which will come in 32B, 122B and 256B sizes, including a smaller 32B version.
AI Companies Brace for New Scrutiny After Democrats' Midterm WinsDemocratic midterm victories are set to intensify congressional scrutiny of tech firms, boost AI safety regulation and tighten local limits on data center construction. Meta's president questioned whether hyperscalers must build outside the US, while Elon Musk touted space-based data centers. Half of proposed US AI data center additions could face tougher local review.
NVIDIA Vera CPU claims top throughput and agent speedNVIDIA said its Vera CPU outperformed both the fastest and most energy-efficient x86 CPUs in a 100-megawatt data center comparison, supporting 1.6x more agents per watt. Early adopters including Perplexity and DeepInfra reported 1.5x to 2.2x speedups on agentic workloads.
Cult.fit uses Codex to boost operational efficiencyCult.fit is using OpenAI's Codex to help teams across functions build tools, move faster and change how work gets done. The case reflects coding agents moving into everyday enterprise operations.
Japan presses US tech giants on security and AIAfter October attacks on websites in several countries, Japan is seeking more software vulnerability information from US tech and AI firms and preparing an emergency cybersecurity package with advance notice of major flaws. PM Sanae Takaichi ordered stronger defenses, and Japan plans AI-simulated cyberattacks.
Millions of dollars of Nvidia chips reportedly stolen in OaklandNvidia chips worth millions of dollars were reportedly stolen in Oakland, highlighting logistics and security risks around high-value AI computing hardware.
AI Fears Rattle Wall Street, Boosting Safe HavensBloomberg's program roundup discusses market-moving news tied to AI-driven fears, featuring fixed-income portfolio managers and strategists. No specific figures are given; the segment reflects investor sentiment and a shift toward safe havens.
NVIDIA releases new Cosmos 3 open models for interactive world simulationNVIDIA added four open-weight models to its Cosmos world model platform for generating interactive driving and robotics scenes. The checkpoints are post-trained from Cosmos 3 Nano with 16B trainable parameters and use the OpenMDW 1.1 license, permitting commercial and non-commercial use.
Hugging Face Releases MoE Training Handbook Built on Megatron-LMHugging Face's Nouamane released a Mixture-of-Experts training handbook covering Megatron-LM settings, from hardware checks and dense baselines to MoE layer details and serving/RL, with 55 figures and scripts released. Lessons draw on about 800k GH200 GPU-hours of MoE training on the Alps cluster at CSCS.
Meta warns data center backlash could hinder AI expansionMeta president Dina Powell McCormick said post-midterm political opposition to data centers could make AI infrastructure investments harder, adding the industry has not done enough to explain the benefits. She said data centers are the visible manifestation of AI and that underlying sentiment about AI is the deeper issue. Meta has data centers in 23 states.
Meta adds real-time voice mode to Muse AI agentMeta CEO Mark Zuckerberg announced that the Muse AI agent now supports real-time voice conversations, rolled out on the iOS app. The avatar animates in sync with speech, and users can change voice tone by asking or through settings. It responded quickly in tests but occasionally paused while looking up information.
Google to Spend $17 Billion on Wyoming Data CenterAlphabet's Google will spend $17 billion to build a data center in Wyoming, putting a price tag on the project just two days after US voters broadly rebuked such massive AI developments.
Analyst: data center pushback may bring financial disciplineGrenadilla Advisory CEO Anna Rathbun told Bloomberg she sees a buying opportunity in less-levered AI-related firms amid the tech selloff, arguing that new oversight from a Democratic-controlled Congress could impose financial discipline on the data center buildout.
FABRIC workshop seeks emergency reviewersThe FABRIC workshop at WACV 2027 is seeking emergency reviewers with expertise in computer vision, multimodal learning, healthcare AI and related fields due to a tight timeline, with reviews due by noon on November 6.
Duke Energy still forecasts data center growth in its regionsDuke Energy Corp. is forecasting a data-center boom in the regions it serves even as backlash threatens AI facility growth elsewhere in the US.
Delta Inference Engine v0.93 boosts local DeepSeek-V4.1 runsDelta Inference Engine v0.93 supports 64–192GB RAM; the Unsloth IQ2 quant of DeepSeek-V4.1-Flash runs at 41 tok/s on one RTX 3090 with 192GB RAM, first token about 40% sooner, scoring 57/99 versus 59 for the full model. It adds Unsloth quants, EXL3 and basic multi-GPU support.
Strategist: AI selloff may deepen on mechanical sellingPeter Tchir, head of macro strategy at Academy Securities, says mechanical selling and related factors could deepen the selloff in the AI trade, and he expects the US economy to slow without increased compute spending.
OpenAI president says company could be IPO ready in 2027OpenAI President Greg Brockman said at a WSJ tech conference that the company could be ready for an IPO in 2027. He cited strong enterprise growth, with annualized revenue up 70% quarter over quarter in Q3 and enterprise now the majority of revenue, up from about 30%.
US stocks plunge after midterms, AI sector hit hardUS stocks fell sharply after the midterm elections, with the Dow down about 1,200 points and the Nasdaq dropping 4.8% into correction territory. Investors worried about regulatory risks from newly empowered Democrats critical of AI data centers; Nvidia fell 7%, AMD 12%, and SanDisk over 15%.
Group says Anthropic accelerating AI development despite slowdown callsIndependent monitoring group METR said Anthropic is developing AI systems about twice as fast as a year ago, even as CEO Dario Amodei called for the industry to slow capability improvements. Anthropic co-founder Sam McCandlish said many of METR's conclusions match his own and that the company will publish its own analysis in coming weeks.
Anthropic teams with Plaid on AI personal finance guidanceAnthropic is partnering with fintech Plaid to use its broad consumer finance data network to give Claude users more tailored financial guidance, extending the model into personal finance use cases.
Claude adds Personal Finance tool to analyze spendingAnthropic launched Personal Finance for Claude, letting users connect bank and card accounts via Plaid and ask questions about spending in plain language. Access is read-only, with no ads or own financial products, as Claude catches up to ChatGPT's similar May feature.
Baidu open-sources 280B-parameter ERNIE 5.5 Flash Base modelBaidu released ERNIE 5.5 Flash Base, an open-source foundation model with 280 billion parameters in a mixture-of-experts architecture, pre-trained on 33.5 trillion tokens. It ships without fine-tuning so developers can customize it freely.
Democratic wave puts AI regulation in focusDemocrats took control of the House and Senate, with holding AI companies accountable a recurring campaign theme. The AI industry now faces the prospect of tighter regulation and slower data center construction.
Data-center backlash may reduce risk of AI investment bustThe backlash against data centers seen in the US midterm elections poses new risks to a key growth driver, but restrictions on development could ultimately reduce the odds of a downturn by keeping the AI investment boom in check.
Nasdaq 100 plunges as tech rout deepensThe tech rout accelerated, slamming chipmakers and sending Nvidia tumbling as the Nasdaq 100 plunged. Forrester founder George Colony discussed AI's next phase, arguing custom chips could challenge Nvidia and that the biggest value may shift to software.
Anthropic CEO calls for stronger AI regulationIn an ABC News interview, Dario Amodei said government should have narrow power to block deployment of unsafe AI models, and announced a $200M investment in research on AI's societal impact. He also floated wage reinsurance, retention incentives and unemployment insurance reform to cushion job disruption.
Rivet launches S3-based agent orchestration platformY Combinator W23 startup Rivet launched an orchestrator for agentic workloads that uses lightweight actors backed by S3 instead of heavy sandboxes: 72 KB baseline memory and 4.9 ms p99 startup versus 512 MB and 1.7 s for a minimal sandbox. It is open source and self-hostable; Amp migrated in a weekend.
The A.I. Daily