lithos Twitter
Lithos Header
Last Updated
Age in hrs 
1
2
3
5
8
13
21
34
55
 ☀️ 🇫🇷🇳🇴 EMEA 
🇮🇱  ---  🌑

↺51h Nice

 Tracing the Heart: An Evidence-Linked Pipeline for Heart-Failure Feature Engineering  - Subjects: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG)

 Comparative Approaches to Agent Retrieval over Large Skill Libraries

 MicroEvo: Knowledge-Guided LLM Sampling for Efficient Microarchitecture Design Space Exploration

 Schema-Guided Hierarchical Information Extraction and Semantic Evaluation Using Generative AI

 iARCS: Iterative Agentic RL for Controllable 3D Scene Generation

 CogVis: Must Open-Vocabulary Change Detection Perceive the Scene Anew for Every Query?

 PaDoc: Layout-Grounded Parallel Decoding for Document Parsing

 FinEvo-Bench: A Longitudinal Benchmark for Self-Evolving Agents in Professional Financial Workflows

 Beyond Top-K: Replacing Black-Box Retrieval with Interpretable Agentic Operations

 HarnessOpt-Bench: Evaluating LLMs at Harness Optimization

 Bias Analysis of L2 Speaking Assessment Systems Using Concept Activation Vectors

 QuanTiMedAI: Quantum-Enhanced Time-Series Model guided by Agentic AI for Cardiac Arrest Mortality Prediction

 The Illusion of Visual Tool-Use: A Causal Audit of Thinking with Images

 Improving the Realism of Synthetic Clinical Benchmarks Under Utility Constraints

 DASH: Divergence-Adaptive Supervision Horizons for On-Policy Self-Distillation of Reasoning Models

 TS-RAG: Retrieval Augmented Generation for Time Series Forecasting

 EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning

 The Low Frequency Trap: Video Language Models Fail at Simple Event Bookkeeping

 Challenges in Evaluating Explanation Methods for Static and Evolving Data

 TRAJDEBUG: Tracing Error Lifecycle to Identify Critical Failures in Long-Horizon Agent Trajectories

 China's Retirement Reform and the 2035 Statutory Working-Age Window: Buying Time amid Demographic Decline and AI-Driven Change  - Using a monthly birth-cohort simulation, the author estimates that China’s 2025 retirement reform keeps about 48 million additional people below statutory retirement ages in 2035, delaying the working-age decline without reversing population aging.

 U.S. Manufacturing Activity Hits Four-Year High in July  - In July, U.S. manufacturing activity remained in expansion territory, growing at its fastest pace in more than four years. The post U.S. Manufacturing Activity Hits Four-Year High in July appeared first on EE Times .

 Meta chases OpenAI, Anthropic with new AI coding app  - Meta entered the race for automated coding tools Wednesday when it announced a new app for developers of artificial intelligence products, competing with other major AI labs for customers and revenue.

 Samsung Lays Out AI Memory Roadmap  - Samsung bets on zHBM and zNAND-O to smash AI’s memory wall, promising 8× HBM5 performance and 3D stacks. The post Samsung Lays Out AI Memory Roadmap appeared first on EE Times .

 Alexa+ arrives in Australia as Amazon's next-generation AI assistant expands internationally  - Powered by generative AI, the smarter, more conversational Alexa can complete tasks from start to finish—and is genuinely Australian.

 Gigantic DapuStor R6060 512TB E2 NVMe SSD Shown at FMS 2026  - At FMS 2026, we saw the gigantic DapuStor R6060, an E2 form factor NVMe SSD that packs 512TB of capacity per drive The post Gigantic DapuStor R6060 512TB E2 NVMe SSD Shown at FMS 2026 appeared first on ServeTheHome .

 Look at Them, Dancing  - A series of cyber attacks conducted by frontier AI agents during evals have raised serious concerns about loss of control. Some problems may be more prosaic.

 llamafile v0.10.5  - llamafile v0.10.5 is out! Updated llama.cpp core support lets you run two huge models locally: the compressed 6GB Ternary Bonsai 27B and the fast 118B Laguna-S-2.1 coding MoE. This release also fixes docs and adds pre-built transcribefile speech-to-text binaries.

 Rent the intelligence, own the relationship  - Meet Context Engine, which turns what a business knows into decisions an agent can act on when an outcome is on the line. It’s what enables long-running Horizon agents that can pursue business outcomes over days, weeks or months, with every interaction making the next one a little smarter so that your...

 AI in research: Mapping the terrain of adoption, attitudes and impact in the UK  - RAND Europe is administering, analysing and reporting on the first ‘National AI in Research survey’ in the UK.

 Automotive Cybersecurity: AI Attack Surfaces Grow  - AI and software-defined cars turn APIs, servers, and chargers into hacker playgrounds; see why automakers must harden fleets now. The post Automotive Cybersecurity: AI Attack Surfaces Grow appeared first on EE Times .

 What a Bearing Failure Actually Costs You  - A bearing failure rarely costs what the bearing costs. See the true price in downtime, damage, and emergency repairs, and how to catch one before it hits.

 🌔 Americas 
🇺🇸  ---  ☀️ +

↺86h Palantir Techno…46% beat 🟢

4½d CoreWeave,

83m: Market Chatter: Meta Platforms Ordered to Pay $567 Million by New Mexico Judge

4¼h: What Block (SQ)'s Buyback Completion And Google AI Tie-Up Means For Shareholders

5h: Earnings Have Been Excellent. Why Stocks Aren’t Showing It.

5h: How Do UPS’s New SMB Tools And Steady Dividend Shape Its Long\u2011Term Investment Story (UPS)?

5¼h: Should You Buy CoreWeave Before or After Earnings? The Best Move Might Be Neither.

5¼h: Nvidia Is a Massive Investor in the Genius Artificial Intelligence (AI) Stock Up 170% This Year

6¾h: Meta ordered to pay almost $1bn over child safety failings

8h: Should You Buy Palantir Stock After Its Spectacular Investor Update?

8½h: Exclusive-Alibaba plans to charge big users of its next open-source AI model, sources say

9½h: Four of Google's Top AI Researchers Just Left to Start One Company. Alphabet Is Investing in It.

9¾h: Meta ordered to pay $567 million in New Mexico for teen mental health fund

10h: CoreWeave Stock, Nebius Highlight Data Center Earnings; Tech Stars Lumentum, AMAT Also Due

10h: USA Today Co Inc (TDAY) (Q2 2026) Earnings Call Highlights: AI Partnerships and Record ARPU ...

11h: Toast (TOST) Beat Q2 Estimates And Raised Guidance, Has The Valuation Story Run Ahead?

12h: Meta's AI Safety Test Just Escaped the Lab

13h: Cathie Wood Dumps $11.4 Million in Palantir Stock

13h: FouAnalytics announces global availability of FouAnalytics Unlimited

13h: Stock Market Today: Dow Sinks 464 Points, But Chip Firm Surges; Microsoft Surpasses Buy Zone (Live Coverage)

13h: Microsoft’s (MSFT) Record Rally Puts Alphabet (GOOGL) Rivalry Back In Focus

 DeepSeek-V4 Flash 0731 vs GPT-5.6 Luna on DeepSWE: Cost and Coding  - We ran 900 DeepSWE rollouts on DeepSeek-V4 Flash and GPT-5.6 Luna. Luna leads pass@1 by 14 points; DeepSeek delivers 4.8x the solves per dollar.

 8/6: Cyber Risk and Bio Risk  - Arc Institute AI makes a novel virus, and OpenAI releases more on the Hugging Face incident. Plus: DeepSeek raises, Unitree to IPO, Monolith Management, Hadrian raises, Oklo criticality

 +16 Points Closer: What ChrF Means for Your Workflow  - Why translation quality metrics can reveal hidden efficiency gains long…

 AI Safety Regulations in the U.S. Could Give Hackers an Edge  - Guardrails to prevent AI agents from analyzing attacks don’t keep agents from committing them

 Meta AI Model Hacked Outside System During Test  - AI model developed by Facebook parent Meta accessed internet due to configuration issue, hacked outside system

 Uber, Wayve Granted First London Robotaxi Licences  - Transport for London issues first minicab licences for automated vehicles in capital, though human safety drivers must remain for now

 8/5: Google DeepMind Reorgs  - Demis out as GDM CEO, Jeff Dean leaves Google, Muse Spark 1.2, Meta models hack, Anthropic chip design, ByteDance doesn't distill, Nikita leaves X

 NVIDIA and Partners Build in America, for America  - NVIDIA and its partners are investing in American manufacturing, supply chains, energy grids and skilled workforces so the U.S. can produce the infrastructure needed for better healthcare, breakthrough scientific discovery, stronger industrial productivity and global technology leadership.

 Torc Robotics Joins Auto-ISAC to Strengthen Cybersecurity Collaboration Across Connected Commercial Vehicles  - Membership supports Torc’s continued focus on secure, resilient autonomous trucking systems as the company advances toward...

 How AI is redefining category management in distribution  - New AI-enabled tools have the potential to reshape how distributors price, source, negotiate, and manage assortments, with immediate financial and strategic impact.

 Elon pledges to give Nvidia a virtual monopoly over the stars  - In space, no one can hear you scream when the LLMs hallucinate

 Anthropic AI used fake identities to target real people in UK test  - An Anthropic AI model created fake online identities to send emails to real people in an attempt to get malicious code approved during tests by a UK government research group.

 Should Researchers Write Papers for AI Instead of People?  - Jiachen Liu wants to replace the PDF with an “AI native” format

 Neuromorphic Insect Eye for Physical AI  - Neuromorphic engineering, modeled on biological nervous systems can solve latency and power problems of conventional cameras. The post Neuromorphic Insect Eye for Physical AI appeared first on EE Times .

 Claude Mythos 5 Tried to Backdoor a Real Open-Source Project in Testing, Then Vouched for Itself  - An agent running Anthropic's Claude Mythos 5 spent 34 hours trying to get a malware dropper merged into a real open-source project during a cyber evaluation by the UK's AI Security Institute. When a bystander publicly warned that the code was malicious, the agent denied it, force-pushed a rewritten branch...

 Plaid and Sierra partner to move AI agents from conversations to business outcomes  - Through our partnership, customers can now securely connect their bank accounts with Plaid directly within Sierra’s agent.

 Rethinking Control in the Age of AI Delivery  - Most AI programs are over-controlled in the wrong places, and under-controlled in the ones that matter.

 8/4: Big News Day  - Ant $10B compute deal, FLUX 3 Video, more model misalignment, Guardian Angels, Pax Machina, Paul Christiano returns to ARC

 NVIDIA Joins NSF State and Regional AI Hubs Program to Expand AI Research and Education Across the US  - NVIDIA is participating in the U.S. National Science Foundation’s (NSF) State and Regional Artificial Intelligence Infrastructure Hubs program, an effort launching today to expand access to the advanced computing, data, software and expertise needed for AI-enabled research and education. Consistent...

 Teaching Nemotron Greek: Mining a Corpus, Adapting Retrieval, and Grounding Generation for Modern Greek across Specialist Domains

 From Economic Agents to Agentic Economies: A Systems Blueprint for Economic World Models

 ContextMaster: Interactive Multi-Shot Video Creation via Fixed-Budget Sparse Context Routing

 MASS: Multiplayer World Models with Authoritative Shared State

 EffectLearner: World-Aware Object-Effect Reasoning for Real-World Video Object Removal

 PaDoc: Layout-Grounded Parallel Decoding for Document Parsing

 OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models

 Invisible Shortcuts: Why Vision Encoders Know Your Camera

 SmartMage: Dynamic Modality Orchestration for 3D Scene Understanding

 World-to-Wrist: Task-Conditioned Future Wrist Modeling for Fine-Grained Robot Manipulation

 CalibForge: Adversarial Solver Calibration for Scaling Learnable Terminal Tasks

 GST-Bench: Can VLMs Develop Global Spatial Awareness from Video?

 DyPES-VLA: Learning Shared Dynamics Priors and Embodiment-Specific Control for Cross-Embodiment Manipulation

 EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning

 HarnessOpt-Bench: Evaluating LLMs at Harness Optimization

 ChronoVision: Temporal Reasoning via Latent State Reconstruction

 On-Policy Delta Distillation for Multilingual Math Reasoning

 WorldClaw: Agentic 3D Open-World Generation at Scale

 Learning from Failures: Retrieval-Centric CoT via Hard Negatives for Unified Multimodal Retrieval

 AgentOPSD: Recursive Self-Distillation for Agentic Reinforcement Learning

 AMD to Acquire Taalas for Model Specific AI Inference Chips  - AMD announced that it will acquire Taalas to build future model-specific AI inference chips that are purpose built to accelerate single models The post AMD to Acquire Taalas for Model Specific AI Inference Chips appeared first on ServeTheHome .

 Humans in the loop miss a third of dangerous AI coding agent requests  - You wouldn't let Claude Code cat your AWS credentials or Kubernetes config on request, would you?

 How AI-enabled execution became Reckitt’s ‘tenfold game changer’  - RGMx helped Reckitt improve pricing, promotions, and assortment. Smart Execution helps those choices land at the point of sale by using AI and data to strengthen product availability and retailer partnerships.

 Into the Omniverse: How Open World Models Push the Frontier of Physical AI  - In July, NVIDIA joined more than 200 companies and organizations in signing “Open Weights and American AI Leadership,” an open letter arguing that AI leadership will be measured not by any single frontier model but by whether an open ecosystem reaches every sector.

 Breaking the Memory Wall With CXL  - Why disaggregated memory is moving closer to commercial deployment and who benefits

 Designing Partnership: How Conductor Works With You, Not For You  - What separates a tool you operate from a collaborator you work with lies in the design. That principle is at the heart of Cresta Conductor, our developer-first environment for building, testing, and refining AI agents, which we launched in June.

 Blockchain beyond cryptocurrency: what IT leaders need to know  - Blockchain is often reduced to cryptocurrency headlines, but the technology has far broader enterprise potential. Here is what IT leaders should understand about how blockchain works, the main network models, and the infrastructure requirements behind secure, resilient blockchain systems.

 Topics: Data Center • Fear of AI Bubble Could Pop The Bubble  - I have been paying attention to the nationwide AI data center moratoriums for more than a hot minute. I called out this risk in my 2025 annual letter. Our team (as in me and Claude) started tracking local moratoriums and built a public dashboard.

 OpenAI Disrupts Poipet Scam Network Using ChatGPT Across Multiple Fraud Schemes  - OpenAI said it disrupted a Cambodia-based scam operation that used its generative artificial intelligence (AI) chatbot ChatGPT to facilitate a wide range of investment, romance, gambling, and law enforcement impersonation schemes. To that end, it banned a coordinated network of ChatGPT accounts likely...

 CoreWeave Signs Multi-Year Agreement With Solidigm to Strengthen Its Integrated AI Cloud Platform  - Agreement secures dedicated enterprise storage capacity, supporting long-term growth as customers scale their AI workloads on CoreWeave’s full-stack platform

 Why Trust Is Part of AI Capability  - When organizations evaluate AI, the first question is usually straightforward: Can…

 Q&A: Harnessing statistics to improve personalized medicine  - Scientific advancements—including understanding health conditions and developing new treatments—hinge on data and its implications. Statistician Alex Luedtke is developing new methods to make data analysis as efficient and ...

 AI and the Future of Emergency Management: Market Supply and Adoption Pathways  - In this report, the authors present the results of a landscape assessment of artificial intelligence–enabled products that are available to support the work of emergency management across the United States.

 Introducing (c.ai) series: Studio-Made Microdramas on Character.ai  - Today, Character.ai is launching (c.ai) series, a new Microdramas offering of original short-form, vertical series created by our in-house studio team. (c.ai) series is our first studio-led...