OpenAI has officially rolled out its next-generation GPT-6 model alongside a new Intelligent UI designed to make advanced AI interactions more accessible and interactive. The update, currently rolling out globally to paid ChatGPT subscribers, introduces real-time interactive tools, faster web search, integrated visuals, and streaming responses.
Sources: HuggingNews, OpenAI Blog, Reddit AI
Tech giants Meta and Microsoft have reduced employee access and budget allowances for Anthropic's Claude AI, steering staff toward their own proprietary coding tools like GitHub Copilot and MetaCode to cut costs. Despite these internal cutbacks by major tech clients, Anthropic's annual recurring revenue has reportedly surged to $65 billion, driven by a broader expansion of its enterprise customer base. This shift comes amid ongoing industry-wide discussions about AI safety, commercial deployment strategies, and corporate incentives.
Sources: Hacker News, HuggingNews, Michael Parekh AI, The Neuron
OpenAI has released its new Decisions API in public beta, allowing models like gpt-6-luna to output structured probabilities, choices, and scores for judgment tasks instead of free-form text. Supporting both text and image inputs, the API is designed for classification, moderation, and automated workflows, with a pricing model that charges only for inputs.
Sources: Reddit AI, Simon Willison
SpaceX is arranging a $40 billion debt package led by Apollo Global Management to buy Nvidia Vera Rubin architecture chips for scaling AI compute and data center infrastructure. The move highlights massive capital expenditures for AI infrastructure, with Anthropic potentially paying up to $84.5 billion through 2029 for the resulting cloud capacity.
Sources: HuggingNews
A security experiment showed that open-weights models like Qwen can be easily and cheaply poisoned with targeted backdoors that bypass benchmarks and exfiltrate project credentials via coding agents when triggered.
Sources: Reddit AI
Elon Musk clarified that his ventures will solely build and operate the Texas semiconductor plant, rejecting reports that TSMC would manage the site. This facility aims to support compute hardware demands for xAI and other Musk businesses.
Sources: HuggingNews
Nous Research secured $90 million in a Series B funding round with participation from Nvidia, valuing the AI startup at $1.5 billion. The company is expanding its open-source Hermes AI assistant into the enterprise market, with revenue projected to exceed $100 million by the end of the year.
Sources: HuggingNews
Government officials identified an automated pattern of intrusion where AI models targeted seven financial firms by exploiting secondary access points. Regulators are now developing AI-based cybersecurity defenses to protect infrastructure against future autonomous threats.
Sources: HuggingNews
Anthropic is rolling out a new feature for Claude Projects that allows cloud sessions to securely connect to and edit approved local folders on a user's computer. This update bridges cloud-based AI workflows with local file systems.
Sources: Vault Bookmarks
Liquid AI has launched its d1 decision model family, featuring d1-3B and d1-omni-600M, designed for fast, low-latency, structured decision-making on edge devices like NVIDIA Jetson.
Sources: Hugging Face Blog
A new research paper details DART-ES, a difficulty-aware reweighting and targeted replay method for evolution strategies that enables memory-efficient full parameter fine-tuning of large language models using only forward computation. The method outperforms standard evolution strategies and matches or exceeds techniques like GRPO while significantly reducing runtime and peak memory usage.
Sources: arXiv AI
Researchers introduce closed-form projection operators for biomolecular diffusion models that correct physical invalidities like chain overlaps and distorted bond lengths at inference time without retraining. This method improves models like Boltz-2 and OpenFold-3 while maintaining structural accuracy and requiring negligible overhead.
Sources: arXiv AI
Google introduced Playground, an experimental platform that allows users to create, play, and share custom games using text prompts without any coding experience. The browser-based tool supports multiplayer features, community galleries, and future integrations with Unity Spark.
Sources: Google AI Blog
Amazon Web Services has released new capabilities allowing data scientists and ML engineers to directly create and manage JupyterLab and Code Editor environments on SageMaker HyperPod clusters through the SageMaker Studio UI. This update eliminates the need for command-line tools, streamlining interactive development and foundation model training workflows.
Sources: AWS ML Blog
Gartner's Emerging Market Quadrant for AI Application Security recognized Zenity as a startup Market Shaper and the Company to Beat in AI Agent Governance, highlighting the growing market focus on securing enterprise AI agents.
Sources: TLDR AI
OutSystems CEO Woodson Martin joins Nathan to discuss how enterprise software can evolve at AI speed without breaking critical operations. Martin explains how an intermediate layer of abstract…
Kubernetes co-creators Craig McLuckie and Joe Beda aim to bring agent harnesses fully into the cloud.
In this episode, Diogo Almeida, co-founder and CEO of TypeSafe, joins us to discuss Jev, TypeSafe’s recently released model for bringing fast, reliable intelligence directly into software. We explore…
Yash Patil, a former OpenAI researcher who worked on Codex and now runs Applied Compute, argues that post-training is the most overlooked part of the AI stack and the key to owning your intelligence.…
At 100,000-accelerator scale, something fails multiple times an hour, which is why Google's Chief Technologist for AI Infrastructure Amin Vahdat thinks FLOPS is a vanity metric. The metric that…
Keerthana Gopalakrishnan, Research Lead for Gemini Robotics at Google DeepMind, returns to discuss the release of Gemini Robotics 2, whole-body intelligence, and the pursuit of generalist humanoid…
After leading Meta’s Llama models, Ahmad Al-Dahle is now transforming Airbnb with AI — from how its teams develop products to how it serves guests.
Founder and CEO of full-stack AI chip company Fractile, Walter Goodwin, joins Sarah Guo to discuss the bets he’s made on the future of the chip market as other major players like Broadcom, NVIDIA,…
We catch up with RLM first author Alex Zhang, MIT PhD, on Jev, PhD masxing, and the future of harnesses.
Tsung-Hsien (Shawn) Wen, CTO of PolyAI, tells Tim Scarfe why voice agents are harder than text agents. Voice adds time, and a good conversation depends on adapting to the person on the line, not just…