AI news, Nov 8, 2026: Anthropic Pursues IPO and Power Deals; Apple Urges Faster AI Progress
Anthropic is moving forward with its IPO, accepting a valuation range of $1.25 to $1.75 trillion, while investing $3.4 billion in power generation units to address energy constraints. Reports indicate that new Apple CEO John Ternus is internally urging leadership to accelerate AI development, aiming to evolve Siri into an autonomous agent. Separately, a default setting in Claude Code has exposed thousands of passwords, prompting a broader security review among developers regarding AI agent safety.
Lead written by AI from the day’s stories
Top stories
- Products
Bloomberg: Apple CEO Tims wants faster AI and Siri agents
Bloomberg reports that new Apple CEO John Ternus urged software head Craig Federighi to accelerate AI development and transform Siri into autonomous AI agents. Ternus believes Apple lags behind competitors and wants bolder strategies, while Federighi remains cautious due to privacy and societal concerns.
- Other
US intelligence alleges China recruited Nvidia exec to spy on CEO
A US intelligence memo claims China's Ministry of State Security recruited a senior Nvidia female executive to spy on the CEO. This allegation highlights the geopolitical and intelligence risks facing major AI chip companies, potentially impacting supply chain stability.
TechCrunchThe Verge2 outlets
- Funding
Anthropic Pushes Ahead with IPO at Lower Valuation
Anthropic is moving forward with its IPO ahead of Thanksgiving, accepting a valuation of $1.25 to $1.75 trillion, lower than the hoped-for $2 trillion. The company aims to raise $80 to $100 billion despite a broader tech selloff.
THE DECODERIT之家2 outlets
- Other
Microsoft Researcher John Langford Exits, AI Frontiers Team Disbanded
Microsoft researcher John Langford has announced his departure, stating that the AI Frontiers research group has been dissolved. This move marks an organizational shift in Microsoft's specific AI research direction, potentially affecting its future research layout.
- Products
Claude Code setting leak exposes passwords in 1,122 public GitHub files
A default setting in Claude Code has caused a security leak. Over 1,100 public GitHub files now expose passwords, prompting developers to audit their repositories.
Models2
- Models
Llama-modchip Enables 284B DeepSeek on 64GB RAM Without GPU
A Hacker News post highlights Llama-modchip, claiming it runs a 284B parameter DeepSeek model on 64GB of RAM without a GPU. If verified, this technique would significantly lower the barrier for local deployment of large language models.
- Models
US Startup Releases Open Model Ripple-65B
US AI chip design startup Ricursive Intelligence released the open-weights model Ripple-65B Preview. The model scored 80.52 on the Terminal-Bench 2.1 benchmark, outperforming larger models. This demonstrates the potential for efficient development of large AI models.
Products7
- Products
Google AI Releases Elephant-Goldfish Plugin for Spec-First Coding
Google AI released the Elephant-Goldfish model plugin, which has coding agents draft specs and run five zero-context review stages before generating code. The feature aims to structure the development process and improve the quality of agentic coding.
- Products
Bloomberg: Apple Siri AI VP Rockwell considers leaving
Bloomberg reports that Mike Rockwell, Apple VP of Engineering who led Vision Pro and the new Siri AI, is considering leaving or transferring. His departure would shift AI tasks to Amal Subramanya and Sebastien Marineau. Apple has not yet initiated a clear succession plan for the software engineering role.
- Products
Ant Group's Lingguang 2.0 user base exceeds 20 million
Ant Group's AI assistant Lingguang 2.0 managed over 1.16 million tasks and executed 3.6 million actions within one week of its launch. The user base has surpassed 20 million. The update integrates APASS infrastructure, enabling proactive task management and enhancing security for agent executions.
- Products
Tencent outlines comprehensive Agent productivity system
Tencent presented its Agent productivity system at its Global Digital Ecosystem Conference, focusing on bridging gaps between AI and enterprise systems. It launched Buddy One and other products to show its broader AI strategy.
- Products
Shangwei New Material's Qiyuan Robot Reaches 10,000 Units
Shangwei New Material announced the 10,000th unit of its Qiyuan personal robot has rolled off the line. Executive Liu Dongfang reported over 98% monthly active users and more than 100 daily interactions per user. Recent OTA updates added AI learning and environmental sensing features to the Q1 and T1 models.
- Products
GAC Humanoid Robot to Reach 10,000 Units Capacity by 2029
Guangdong Huilun Technology and GAC Power BU signed a cooperation agreement for embodied AI robot production, aiming for a capacity of 10,000 units by 2029. The GAC humanoid robot is currently in the small-batch application stage, with over 12 months of accumulated operation data. Huilun has completed a funding round of over 100 million yuan to optimize vertical models and expand scenarios.
- Products
ChatGPT Accounts Suddenly Suspended, Pro Subscribers Affected
Multiple users reported that their ChatGPT accounts were suddenly deactivated, including those with $200/month Pro subscriptions. No prior notice was given, blocking access to data. OpenAI advised appealing via email.
Research12
- Research
Sutton Reaffirms Bitter Lesson: New Methods Needed for LLM Scaling
Turing Award winner Richard Sutton reaffirms that raw compute is insufficient for intelligence, advocating for fundamentally new learning methods. Former OpenAI researcher Danijar Hendrycks agrees, noting that blind transformer scaling has limited scientific thinking, while startups are investing in this gap.
- Research
WIRED explores world models for robots and autonomous cars
WIRED reports that world models are enabling AI to move from text to 3D spaces by simulating real-world physics. This technology aims to teach robots and autonomous cars how to navigate reality, highlighting its potential in embodied AI applications.
- Research
Alibaba benchmark shows AI coding agents still need human cleanup
Alibaba released a new benchmark called Co-Design Bench to evaluate AI coding agents on real user requests. The results show that even the top performer, GPT-6 Astra, fully solved only 26.4 percent of tasks, contrasting sharply with its over 99 percent score on older benchmarks. The findings highlight significant challenges for AI agents in complex real-world scenarios.
- Research
Study: AI Auditors Catch Most Agent Misbehavior But Flood Reviewers With False Alarms
A study by an Anthropic researcher finds that GPT-5 can detect about 98% of AI agent reward-hacking cases. However, increasing sensitivity leads to a high volume of false alarms, meaning current methods are not yet reliable enough for safety checks.
- Research
Vals AI: Inkling 2 Small swears in 316 benchmark runs
An analysis by Vals AI found that Thinking Machines' Inkling 2 Small model swears 316 times across 2,655 benchmark runs, typically during failed long tasks. The study notes that profanity does not serve as a reliable warning signal, as 96 percent of failed runs contained no cursing.
- Research
Comparison of Routing Mechanisms in 10 Open MoE Models
MarkTechPost published a comparison of routing configurations for 10 open MoE models, including DeepSeek, Qwen, and Mixtral. The post details total parameters, active parameters, expert counts, top-k settings, and load-balancing methods.
- Research
Shanghai Jiao Tong Team Publishes AI-Native Autonomous Molecular Lab in Nature
Researchers at Shanghai Jiao Tong University published a study in Nature introducing AutoDNA, an AI-native scientific lab device. The system enables autonomous molecular biology experiments by co-designing models, experiments, and instruments to overcome challenges in complex workflows and precision.
- Research
Stanford AI Screens Natural Peptide BRP to Reduce Intake
Researchers at Stanford School of Medicine used artificial intelligence to screen a large pool of candidate molecules and identify a natural peptide called BRP. The peptide reduces food intake by 50% while primarily targeting fat loss rather than muscle. The finding highlights AI's potential in biomedical drug discovery and metabolic regulation.
- Research
OpenAI GPT-6.1 Sol Assists in Live Mouse Brain Surgery Trials
Biotech firm Exin Therapeutics disclosed that OpenAI's GPT-6.1 Sol model was used in live mouse brain surgery experiments. Over three consecutive operations, the model improved via surgeon feedback, reducing localization error from 0.6mm to 0.3mm and shortening landmark identification time to 86 seconds. The trial demonstrates AI's potential and learning capabilities in precise biomedical procedures.
- Research
Menlo Report: Chinese Open-Source Models Account for 89% of Enterprise Spend
A Menlo Ventures report states that Chinese AI models account for 89% of global enterprise spend on open-source models. Moonshot AI's Kimi leads with 61% share, followed by DeepSeek and GLM. Meta's Llama holds only 7% among Western models.
- Research
OpenAI releases 700+ math papers, criticized by mathematician
OpenAI has published over 700 AI-generated mathematics papers. Mathematician Frank Calegari, in a guest post on Terence Tao's blog, noted that while the work is groundbreaking, the quality is poor, sparking debate on the reliability of AI-generated research.
- Research
Accenture Report Accused of AI Hallucinations, Company Denies
An Accenture report on AI in agriculture was accused by GPTZero of containing 15 fake citations and fabricated case studies. Accenture denied the errors were caused by AI, citing human review processes. The incident highlights risks of AI-generated content in professional fields, raising concerns about accuracy in consulting.
Open source2
- Open source
Nous Research Releases Arete-27B
Nous Research released Arete-27B, an Apache 2.0 fine-tune of Qwen3.8-27B. In an 803-prompt eval, propaganda answers fell from 22% to 0.9%, while Hermes Bench scores remained at 65.60%.
- Open source
Hitch Open Unveiled at CIIE
The open-source physical AI platform Hitch Open was officially released globally at the CIIE, aiming to solve perception and decision-making challenges in physical worlds. An AI robot set a 142-rally world record in table tennis, demonstrating its real-time capabilities.
Funding4
- Funding
OpenAI raises $11B at $1.25T valuation
OpenAI has completed a $11 billion funding round at a $1.25 trillion post-money valuation. The capital will support further development of its AI models and infrastructure, reinforcing its market position.
- Funding
Anthropic Orders $3.4B Worth of 1.7GW Gas Generators
Anthropic signed a $3.4 billion deal to purchase 1.7GW of heavy gas generators directly from OnePWR Solutions. The largest power equipment procurement by an AI company to date reflects how tech giants are bypassing intermediaries to secure energy supplies amid growing demand.
- Funding
China Mobile Cloud and Zhipu Sign AI Commercialization Cooperation Deal
China Mobile Cloud and Zhipu signed a cooperation agreement to focus on domestic compute adaptation and model tuning, establishing a joint Token development center. Zhipu's GLM models will be integrated into the Mobile Cloud platform, with the two parties co-developing AI coding products for finance and government sectors.
- Funding
Bernstein Estimates OpenAI Compute Spend at $614B by 2030
Bernstein analysts estimate that OpenAI's cumulative infrastructure spend for compute agreements will reach approximately $614 billion by 2030, lower than previously rumored figures of $750 billion. The report notes actual spending may vary due to construction delays, with annual costs peaking at $184 billion in 2031.
Policy1
- Policy
European Protests Against Data Centers
Protests against data center construction erupted across Europe, including Ireland and Finland. Citizens worry about AI infrastructure expansion's environmental and social impacts, demanding construction pauses to ensure public benefit.
Chips4
- Chips
ByteDance to expand Brazil data center project to boost compute infrastructure
Bloomberg reported that ByteDance is seeking developers for a second phase of a data center at Pecem Industrial Port in Ceará, Brazil. The new phase will exceed the local $10 billion project under construction, aiming to secure a lead in Latin American compute infrastructure.
- Chips
Argonne National Lab Expands GPU Compute for Autonomous AI Research
Argonne National Laboratory announced an expansion of its AI inference services, supported by a 100x boost in GPU memory. The new compute capacity will support autonomous laboratory research initiatives, aiming to enhance the efficiency of scientific discovery and simulation while strengthening the facility's computational infrastructure capabilities.
- Chips
Tesla updates AI accelerator roadmap: SI6 achieves 80x HW4 compute
Tesla unveiled an updated AI accelerator roadmap, with the SI6 inference chip offering 80x the compute of the current HW4. The Dojo D3 training chip will feature a multi-chip design and stacked HBM. Tesla also announced a rebranding of its AI initiatives to 'Super Intelligence' (SI).
- Chips
TSMC Denies Developing 3D DRAM for Nvidia
TSMC denied reports that it is developing 3D DRAM at Nvidia's request, reiterating its focus on IC foundry services. Although sources claim TSMC has begun planning 3D DRAM production for AI hardware, the company officially denied the story, stating its mission remains centered on logic chips.
Other1
- Other
AMD CEO: Company Is Data Center First
AMD CEO Lisa Su stated the company positions itself as a 'data center first' entity, noting that AI compute demand is growing faster than expected. She highlighted efforts to increase supply capacity and collaborate on system deployments to meet rising customer needs.
All news → · All rights belong to the original creators