OpenAI frontier models and Codex are now available on AWS
(openai.com)
AI
OpenAI says its frontier models and Codex are now generally available on AWS, delivered through Amazon Bedrock using AWS-native security and governance controls to reduce friction for deploying the tools in production. It also notes plans to expand OpenAI capabilities on AWS, including future availability of Daybreak (cyber-focused models and Codex Security) to help defenders with earlier risk detection and safer development workflows.
AI is crushing a generation of startups built before ChatGPT
(cnbc.com)
AI
CNBC reports that valuations for U.S. “pre-ChatGPT” unicorn startups have fallen sharply as generative AI shifted investor capital toward AI-native firms, leaving many older companies with stale valuations and outdated technology; PitchBook data cited shows nearly half of America’s 857 unicorns haven’t raised new funding in three years and that startups last valued in 2021 and 2022 are down 68% and 52% on average, with more than 220 firms classified as “fallen unicorns.”
Michael Burry Just Called Nvidia's SpaceX Chip Deal 'Fugazi.'
(247wallst.com)
AI
24/7 Wall St. reports that investor Michael Burry criticized Nvidia’s arrangement with xAI—where Nvidia sells GB200 GPUs into a financing/lease structure involving Valor Compute Infrastructure and Apollo, and then books revenue up front—calling it “fugazi” due to what he says is obscured, multi-layered credit risk potentially tied to annuity holders and leveraged, hard-to-value assets.
AI in SRE: Where and how Google is deploying agentic AI to improve operations
(cloud.google.com)
AI
Google’s SRE team describes “SRE AI,” an initiative to deploy agentic AI across the software development lifecycle—using agents for reliability design, anomaly detection/alerting, incident orchestration and postmortems, and incident investigation/mitigation—while emphasizing transparency, governance, security/privacy, and reliability SLOs over fully black-box automation.
Visa invests in Replit to power agentic payments for developers
(techcrunch.com)
AI
Visa has made an undisclosed investment in AI coding platform Replit and is exploring ways to integrate Visa’s payment offerings into Replit so developers—and the AI agents they build—can accept payments within the platform, including using Visa’s Trusted Agent Protocol. The companies say the work is still exploratory and no joint product has been formally announced.
Build a Basic AI Agent from Scratch: Tools
(ruxu.dev)
AI
The article explains how to make a more capable “AI agent” by giving a language model tool-calling capabilities, then walks through implementing core tools such as running bash commands, reading/writing and editing files, searching files with glob/grep, and fetching web pages as plain text with size limits.
Hating AI is good, actually
(thehandbasket.co)
AI
The Handbasket argues that “hating AI” should be taken seriously, citing examples of AI producing or propagating errors and raising questions about trust in research, publishing, and even literature, while describing AI backlash among professionals and students amid pressure to adopt AI.
Florida AG files lawsuit against OpenAI, CEO Sam Altman for deceptive practices
(myfloridalegal.com)
AI
Florida Attorney General James Uthmeier announced a first-in-the-nation state-led civil lawsuit against OpenAI and CEO Sam Altman, alleging deceptive practices involving ChatGPT—including claims that the product was marketed to children while concealing safety risks and suppressing internal warnings.
Qwen3.7-Plus: Multimodal Agent Intelligence
(qwen.ai)
AI
The article titled “Qwen3.7-Plus: Multimodal Agent Intelligence” discusses Qwen3.7-Plus as a multimodal agent-focused intelligence model, though the provided article text is too limited to extract specific features or claims.
Anthropic confidentially files to go public
(cnn.com)
AI
Anthropic has confidentially filed with the SEC to pursue an initial public offering, though it has not yet set how many shares it would sell or at what price. The filing comes after the company raised $65 billion and valued it at $965 billion, and would provide investors a first detailed view of its financials amid ongoing discussion of AI-market risks. The story also notes Anthropic’s recent competition with OpenAI and Google, alongside regulatory and security concerns tied to its technology.
Macron announces 93B euros in 'Choose France' investments
(euractiv.com)
AI
Emmanuel Macron said at an international conference that he expects confirmed “Choose France” investments of 93 billion euros, including projects for artificial intelligence, data centres and semiconductors, with more than 15,000 jobs planned; he cited 45 billion euros pledged by SoftBank for data centres in northern France by 2031 and said France’s nuclear-powered electricity is a key draw.
Nvidia RTX Spark Laptops
(nvidianews.nvidia.com)
AI
Nvidia and Microsoft announced RTX Spark, a new Blackwell-based Windows PC “superchip” aimed at running secure on-device AI agents, combining up to 1 petaflop of AI performance with up to 128GB unified memory and Windows security primitives plus the NVIDIA OpenShell runtime.
Anthropic confidentially submits draft S-1 to the SEC
(anthropic.com)
AI
Anthropic says it has confidentially submitted a draft Form S-1 to the SEC for a proposed IPO, which would be possible after the SEC’s review, though the share count and price are not yet set and the filing is not an offer to sell securities.
Nvidia unveils general-purpose chip for laptops and desktop PCs
(tomshardware.com)
AI
At Computex 2026, Nvidia announced the RTX Spark Superchip, pitching it as a Windows-on-Arm platform for “agentic” AI PC experiences, supported by up to 20 Arm CPU cores, a Blackwell GPU with 6,144 CUDA cores, and up to 128GB of LPDDR5X unified memory. The company says first systems arriving this fall will include laptops from major OEMs (including a new Surface Ultra) and “10 or so” desktop models, and it also highlights gaming features (up to 100 FPS at 1440p, per Nvidia) plus GPU-accelerated updates planned with Adobe for Photoshop.
Nvidia Cosmos 3
(developer.nvidia.com)
AI
NVIDIA’s Cosmos 3 is presented as an open, unified foundation model for “physical AI” that combines a reasoner (vision-language) tower with a generator (diffusion-based) tower to produce physics-aware future observations and action sequences, targeting robotics, autonomous driving, and warehouse monitoring. The release includes open-sourced model checkpoints, datasets, training/post-training scripts, and deployment options such as NVIDIA NIM microservices with performance optimizations and support for different model sizes (Nano and Super).
MacBook Pro Rival with the Nvidia Powered Surface Laptop Ultra
(windowslatest.com)
AI
Windows Latest reports that Microsoft has unveiled the Surface Laptop Ultra, a high-end Windows on Arm laptop built in partnership with NVIDIA, featuring an “RTX Spark” platform, up to 128GB unified memory, a 15-inch mini-LED touchscreen, and claims of strong local AI performance; the device is expected to arrive in stores in fall 2026.
Nvidia announces new AI chip for personal computers
(bbc.com)
AI
Nvidia unveiled its RTX Spark AI chip for Windows PCs at Computex in Taipei, positioning it for “personal AI agents” and partnering with Microsoft on a secure Windows platform. The chip will be used in new PCs from Lenovo, HP, Dell, Microsoft Surface, Asus and MSI, with availability planned for autumn, while the company’s move is also happening alongside tightened US export rules on advanced chips to Chinese firms.
LLMs Are Closer to Religion Than They Appear
(theregister.com)
AI
The Register opinion argues that large language models function in a way that parallels religion, citing a lengthy Vatican-style AI policy document and a religiously affiliated study about AI’s failure to provide religious answers; it warns that the resulting debates could drive politically motivated pressure to steer LLM training toward specific religious materials under the banner of “fairness.”
Hard-Won Lessons from a Year of Using AI
(spin.atomicobject.com)
AI
An Atomic Object developer reflects on a year of integrating AI into daily work, arguing that real gains come from human judgment at the start and end of tasks, setting realistic productivity expectations (e.g., roughly 2x rather than 10x), and avoiding multitasking that AI can make more tempting. The article emphasizes using AI for the “middle” work (drafting, research, synthesis) while carefully reviewing outputs and managing agent “wait” time to stay engaged with verification and planning.
Is Python Becoming Pinyin?
(lernerpython.com)
AI
After seeing how much of the PyCon US conversation centers on AI “agentic” (vibe) coding, the author argues that Python may be used less for direct application development as agents can output other languages like Rust or Go, raising questions about who maintains the Python ecosystem and who will learn Python. The piece concludes that Python is unlikely to disappear due to its deep adoption, but its role may shift toward teaching core software-engineering concepts and providing high-quality training data for AI agents rather than being the default target language.