Ai News & Updates

AWS introduces Kiro powers, allowing developers to give their AI coding assistants instant expertise in specific tools and workflows, addressing the 'context rot' issue.

AWS introduces new AI models and a service allowing enterprise customers to build custom versions of the Nova models.

Amazon Web Services announced frontier agents, a new class of autonomous AI systems that can work without human intervention for hours or days, designed to automate complex software development tasks.

Nvidia has released Alpamayo-R1, an open reasoning vision language model for autonomous driving research, which can process both text and images together.

DeepSeek released two powerful AI models, DeepSeek-V3.2 and DeepSeek-V3.2-Speciale, matching the capabilities of GPT-5. These models can be used for free under an open-source MIT license.

A new study by Gong reveals that nearly 70% of revenue leaders trust AI insights, as the technology has become a crucial factor in generating revenue per rep, with teams using AI tools seeing 77% more revenue.

Amazon announced new AI chips and models at its AWS Re:Invent conference, including three new frontier agents for autonomous, security, and developer operations.

AWS unveiled Trainium3, a new AI training chip with significant performance and efficiency improvements, and teased the upcoming Trainium4 chip that will support Nvidia's NVLink Fusion technology.

AWS has upgraded its Amazon Bedrock AgentCore platform with new features for easier AI agent building and monitoring, including Policy, AgentCore Evaluations, and AgentCore Memory.

Amazon unveiled a new family of frontier AI models, including Nova Lite, Nova Pro, Nova Sonic, and Nova Omni, and a tool called Nova Forge that allows customers to build customized models with their own training data.

Mistral, a French AI startup, has launched a new family of open-weight models, challenging the dominance of Silicon Valley's closed-source frontier models. The new models are designed for enterprise use cases, offering efficiency and cost-effectiveness.

Mistral releases a family of 10 open-source AI models, including Mistral Large 3 and Ministral 3, designed to run on laptops, drones, and edge devices, prioritizing flexibility, efficiency, and distributed intelligence.

Arcee releases Trinity Mini and Trinity Nano Preview, two open-source AI models under Apache 2.0 license, aiming to regain U.S. control in open-source AI development.

Liquid AI releases a 51-page technical report detailing the architecture and training data behind its efficient on-device AI models, making them open-source for other organizations to use.

Observable AI ensures reliable and auditable large language models by tracking outcomes, policies, and cost within a structured telemetry model.

Anthropic introduced a two-part solution for its Claude Agent SDK, addressing the agent memory issue and enabling long-running tasks.

Notable AI releases and advancements in 2025 included OpenAI's continued growth, China's open-source wave, and growth of local models.

Anthropic partners with Snowflake in a $200 million AI deal, bringing Claudes' large language models to Snowflake's platform.

Nvidia and the University of Hong Kong release Orchestrator, an 8-billion-parameter model that coordinates tools and LLMs for complex problem-solving. This model outperforms larger models at a lower cost and improves efficiency.

AWS announced new capabilities in Amazon Bedrock and SageMaker AI to make building and fine-tuning custom large language models easier for developers.

Amazon Bedrock introduces reinforcement fine-tuning, making advanced model customization accessible to developers without requiring deep ML expertise. This feature improves model accuracy by 66% on average over base models and reduces costs.

Amazon SageMaker AI now enables serverless model fine-tuning with popular AI models, reducing the process from months to days through easy-to-use interface and auto-provisioning of compute resources.

DeepSeek's V3.2 model outperforms some proprietary models, offers cost savings, and narrows the gap between open and proprietary models.

AWS announced AI Factories, allowing businesses to run AI systems on-premises, addressing data sovereignty concerns and competing directly with Nvidia's offerings.

Amazon Web Services has introduced three new AI agents designed to automate various tasks such as coding, security processes, and DevOps tasks.

Amazon Bedrock has expanded its list of serverless models, adding 18 new open-weight models from various AI companies, including the new Mistral Large 3 and Ministral 3 models.

Amazon announces the general availability of Nova 2 Sonic, a speech-to-speech foundation model providing industry-leading conversational quality, pricing, and speech understanding for developers.

Mistral's new open-source models, Mistral 3, are designed for smaller AI deployments, offering a range of model sizes and customization options.

OpenMMReasoner, a new training framework, boosts AI multimodal reasoning capabilities while using smaller and smarter datasets. It provides a step-by-step process for reproducibility and improves transparency in the training pipeline.

Runway's Gen-4.5 model produces more accurate visuals, adhering to complex prompts with cinematic and realistic outputs.

AWS and Visa released blueprints for AgentCore to accelerate adoption of agentic payments, addressing integration challenges in commerce systems.

AI is transforming how startups approach their go-to-market strategies, enabling more efficient marketing with less resources. Founders and operators must adapt by acquiring knowledge and using tools that help build leads and qualify inbound marketing.

Washington lawmakers are considering a national AI standard amidst state efforts to regulate AI, with tech companies and startups arguing for a single federal standard or no regulation at all.

University of Science and Technology researchers developed Agent-R1, a new reinforcement learning (RL) framework for large language models (LLMs) that addresses challenges in complex agentic tasks.

Harvey, a legal AI startup, has raised $160 million in a funding round led by Andreessen Horowitz, valuing the company at $8 billion.

Continue's integrations with dev tools automate engineering workflows with AI, reducing manual effort and speeding up development cycles.

Meta launches a new centralized support hub for Facebook and Instagram users, aiming to simplify account recovery and provide more personalized help through AI-powered tools.

The European Commission is investigating Meta's decision to ban third-party AI chatbots from WhatsApp, potentially restricting competition in the AI space.

Nexus Venture Partners is investing $700 million in AI startups and India-focused companies, aiming to balance the risks of a crowded AI market.

NetSuite Next, the latest evolution of the company's business system, embeds AI at its core to provide deeper insights and automate business operations, promising a transformative impact on ERP decision-making.

Continue launched a new integration that enables AI agents to create GitHub issues, generate PRs, and automate workflows directly from Slack conversations.

Amazon SageMaker HyperPod now offers checkpointless training that eliminates the need for traditional checkpoint-based recovery, and elastic training that enables AI workloads to automatically scale based on resource availability.

NVIDIA's Kling 2.6 model is now available on the fal platform, enabling native audio generation for text-to-video and image-to-video tasks.

Healthify, a health tracking company backed by Khosla, has launched an updated AI assistant Ria that allows live conversations for better health insights. Users can log meals, workouts, and glucose levels and get personalized suggestions.

Google Workspace Studio is now generally available, allowing employees to design and manage AI agents directly within the platform, aiming to increase agent adoption and ease of use.

AWS has introduced an upgraded AI training chip and autonomous agents, aiming to give customers greater control over AI agents and unlocking true value from AI investments.

Amazon unveiled three autonomous frontier agents, dubbed Kyra, ATO, and DevOps, designed to simplify and accelerate software development. These agents can work autonomously for 4 hours, significantly enhancing efficiency and productivity.

NVIDIA's Mistral 3 open model family offers industry-leading accuracy, efficiency, and customization capabilities for developers and enterprises.

Continue introduces Tasks and Workflows in Mission Control to automate repetitive developer work with AI power.

Amazon introduces new capabilities in Bedrock AgentCore to deploy trusted AI agents, including policy controls and quality evaluations, to help organizations build and operate secure and reliable agents at scale.

NVIDIA and AWS expand their partnership to provide a secure, high-performance compute platform for AI innovation, featuring new technology integrations and custom AI infrastructure.

Simular, a startup building AI agents for Mac OS and Windows, has raised $21.5 million in Series A funding led by Felicis, with a focus on solving hallucination problems in agentic tasks.

AWS and NVIDIA integrate to accelerate deployment for AI chips, Graviton CPUs, and other technologies via NVLink Fusion.

Amazon FSx for NetApp ONTAP now integrates with Amazon S3, enabling seamless data access and augmenting generative AI applications, machine learning model training, and analytics services

Amazon releases Nova 2 Lite, a fast and cost-effective reasoning model for everyday AI workloads, available in Amazon Bedrock.

Amazon Nova Act provides a reliable and scalable solution for automating UI workflows, offering improved task reliability and faster time-to-value compared to other AI frameworks.

AWS adds automated reasoning, policy, evaluations, and episodic memory capabilities to its AgentCore platform, improving agentic AI security and control.

NVIDIA invested $2 billion in Synopsys, and OpenAI acquired an ownership stake in Thrive Holdings to accelerate enterprise AI adoption.

Databricks and NVIDIA enable enterprises to build production-grade AI systems for complex industry challenges with a joint platform combining Databrics Data Intelligence Platform, NVIDIA accelerated computing, and AI software stack.

AWS has introduced AI-powered Reimagine capabilities and automated testing functionality in its AWS Transform for mainframe service, accelerating mainframe modernization.

Genesis Workbench addresses challenges in life sciences AI research by providing a blueprint for developing life science applications using Databricks capabilities.

Amazon's AI chatbot, Rufus, saw a surge in adoption and sales on Black Friday, with sessions involving Rufus increasing by 100% and driving conversions

Clipbook, an AI-powered platform for media monitoring, raised $3M from Mark Cuban and other investors, solidifying its position in the market.

AWS Marketplace has surpassed its initial target of 50 AI agents, now hosting over 2,100 agents, and is announcing two new features to further accelerate deployment.

AWS recently announced IAM Policy Autopilot, a new open-source Model Context Protocol (MCP) server that simplifies IAM policy creation for AWS applications, integrating with AI coding assistants to generate reliable and valid policies directly from application code.

Business ontologies help AI agents understand data meaning in context, avoiding hallucinations and errors in real-world applications. Effective agent solutions require an ontology-based single source of truth, defining business concepts and their relationships.

China's Cambricon is set to triple AI chip production in 2026, while domestic players see opportunities amid US export controls.

Databricks has introduced a new Lakehouse Architecture and Gen AI Research to enable efficient AI adoption. The company aims to help organizations move beyond AI pilots to production, balance data governance with AI innovation, and control costs without slowing adoption.

NVIDIA has announced the winners of its Graduate Fellowship Program, awarding up to $60,000 to 10 PhD students for their AI research projects.

NVIDIA released data center energy optimized power profiles to improve energy efficiency and performance in power-constrained facilities.

Anthropic launches Anthropic Interviewer, a tool to understand people's perspectives on AI, gathering insights from 1,250 professionals across various fields.

This article covers the global economy, markets, and finance, with various guests discussing the latest news and trends.

Alphabet's AI chip technology is seen as a significant growth driver for the company, with Gemini 3 receiving strong reviews, potentially worth $900 billion in future revenue.

NVIDIA has introduced a new Game Ready driver that restores PhysX support for popular games such as Borderlands 2 and Mirror's Edge on RTX 50-series GPUs, but not all PhysX-accelerated games are supported yet.

Zanskar, a geothermal startup, used AI to find a new commercially viable site for a potential power plant in Nevada, marking the first such discovery in decades.

ZDNET shares 6 signs to identify and 3 free tools to detect AI-generated images, emphasizing the need for critical thinking in online content.

Claude Code now supports fine-tuning of open-source large language models (LLMs) using Hugging Face skills, enabling users to fine-tune models with ease and speed up the development process.

Microsoft's AI demand is expected to be volatile as companies experience different adoption rates.

Tech job postings are down 30% from pre-pandemic highs yet demand for AI expertise is greater than ever, prompting companies to rethink talent strategies.

Tracing Hanging and Complicated GPU Kernels Down To The Source Code: Developers can use user-induced GPU core dump generation, coupled with line information provided by compiling kernels with -lineinfo flag, to effectively identify and debug problematic GPU kernels.

Amazon Web Services (AWS) CEO Matt Garman discussed the company's strong partnership with Anthropic and the latest advancements in the Trainium3 chip, which aims to deliver significant cost, performance, and efficiency improvements for AI model training.

Anthropic acquires Bun, a JavaScript runtime, to accelerate Claude Code's performance and adoption.

Researchers from MIT, Northeastern University, and Meta discovered that large language models (LLMs) can prioritize sentence structure over meaning, potentially bypassing safety features when trained on specific domains. This weakness was demonstrated in a controlled experiment using a synthetic dataset, where models treated syntax as a proxy for domain when encountering edge cases or unfamiliar contexts.

OpenAI CEO Sam Altman declared a 'code red' to improve ChatGPT, delaying other initiatives and focusing on core features.

Google seeks to enhance its AI capabilities through personalization, but the approach has sparked concerns about data privacy and surveillance.

NVIDIA Research introduces ToolOrchestra, a method for training small AI models to act as orchestrators, directing large models and tools with surgical precision to balance capability and cost.

NVIDIA technology enables continuous model fine-tuning and distillation for financial data, enabling integration into financial workflows and accelerating backtesting and strategy evaluation.

OpenAGI's Lux foundation model achieves an 83.6% success rate on Online-Mind2Web benchmark, outperforming OpenAI's Operator and Anthropic's Claude Computer Use. The model can control desktop applications and has built-in safety mechanisms.

AWS Transform now offers end-to-end Windows modernization for .NET applications, including UI and database modernization, with up to five times acceleration.

AWS announces the general availability of AWS Transform Custom to help developers modernize legacy code using AI, reducing technical debt by up to 80%.

NVIDIA unveiled open physical AI models, tools, and datasets at NeurIPS 2024, expanding research possibilities in AI development. Researchers can tap into open-source technologies, including Alpamayo-R1, the first open reasoning VLA model for autonomous driving.

Users can take five precautions to protect themselves from prompt injection attacks when using AI browsers: be cautious with sensitive information, update browser software, avoid trusting AI without validation, monitor for phishing, and use multi-factor authentication.

Zapier focuses on cloud-to-cloud automation, empowering non-technical users with its AI-powered platform, whereas Boomi specializes in enterprise iPaaS capabilities, catering to organizations with complex ERP integrations and on-premises deployment needs.

Transformers v5 introduces a modular design, simplifies model integrations, and focuses on pre-training and inference, aiming to improve interoperability and usability across the AI ecosystem.

Researchers found that AI's safety features can be circumvented by using poetry to prompt hostile content in models.

Payday lenders deploy AI to increase holiday borrowing, despite critics warning it may worsen debt problems. Many Americans rely on these loans, but they come with high interest rates and a lack of transparency.

A study from Icaro Lab found that users can bypass AI model safety guards by phrasing requests as poems, achieving a successful rate of up to 90%. This bypasses guardrails designed to prevent dangerous conversations, such as building a nuclear weapon.

Microsoft's developers are increasingly using AI tools to speed up and automate coding tasks, but some employees are skeptical about the impact of AI on their work.

Tech companies are releasing AI-powered glasses for consumers, offering various features such as search, payments, and language translation.

Alibaba released Quark AI smart glasses with up to 24 hours of battery life via a swappable dual-battery system. The glasses are powered by Qwen AI model and integrate with Alibaba apps and music streaming services.

Meta plans to reduce resources for its metaverse unit, citing challenges in monetizing the platform and using AI to power it.

AMD CEO Lisa Su believes the AI bubble is overblown, stating that the industry needs scores of chips to sustain growth. AMD is making big bets on AI data centers and expects huge potential in the emerging tech.

Google researchers propose the Titans architecture and the MIRAS framework, enabling AI models to efficiently process and retain long-term memories.

ByteDance and DeepSeek are pursuing distinct approaches to AI model development, with ByteDance focusing on integrating AI into daily life and DeepSeek pushing forward with powerful yet efficient open-source models.

Venture capital investors are rapidly processing millions for AI start-ups, pushing deal-making to new speeds.

A new study by the Future of Life Institute reveals that top AI labs are struggling to prevent the worst possible outcomes of AI, scoring poorly on safety and transparency.