AI
8 items · 20 articles in this topic
Agent products moved closer to money and autonomy, with OpenAI's Wallet and Anthropic's parallel workers in Claude Code. Open-source inference tooling also had a busy day.
- 1
OpenAI launched ChatGPT Wallet, which lets Dots agents make purchases for US Pro subscribers after biometric confirmation. Agents with payment authority widen the blast radius of prompt injection.
- 2
Fortune reports OpenAI is disbanding its robotics team as its general models begin controlling robots.
- 3
Anthropic added hands-free parallel AI worker execution to Claude Code projects. It also published a resource page describing a Claude Meetings feature.
- 4
The ggml project, the tensor library underpinning llama.cpp, released version 1.0.0.
- 5
Unsloth released updated Dynamic v3.1 quantized builds of the Qwen3.8-27B model.
- 6
Cloudflare open-sourced Kitesurf, the stateless web browser it built for AI agents.
NotablePhoronix ↗
- 7
The FAA expanded its use of an AI tool to manage air traffic congestion across all US airspace, without involving the controllers' union.
- 8
AMD used an AI agent workflow with HRX Loom to port 50 Hugging Face models to AMD client GPUs in about three days.
Phoronix ↗1 min read
More in AI · 12 articles›
A forensic report claims Chinese AI labs including DeepSeek and Moonshot trained models on Nvidia Blackwell chips despite export controls.
The Registerexport-controlsnvidiallmFeed excerpt
Forensic report finds Blackwell's fingerprint in DeepSeek, Moonshot AI, MiniMax and Ant Group models
OpenAI launched ChatGPT Wallet, letting Pro subscribers in the US have Dots agents make purchases after biometric confirmation.
The VergeopenaiagentspaymentschatgptFeed excerpt
After launching its Dots AI agents platform earlier this year, OpenAI is launching a wallet for the agents. ChatGPT Wallet is rolling out starting today for ChatGPT Pro subscribers in the US aged 18 and older using a US billing address and "eligible US-issued American Express, Chase, Mastercard and Visa cards." After adding a card to the wallet, your Dots agents can use it to buy things in response to your prompts, but they'll check in with you each time before checking out. OpenAI's support page says every purchase with ChatGPT Wallet requires a security check with biometric verification like Face ID or a fingerprint scan. Now your dot c … Read the full story at The Verge.
Anthropic published a resource page describing Claude Meetings, a Claude feature for meetings.
Hacker News (front page)claudeanthropicllmFeed excerpt
Article URL: https://claude.com/resources/articles/claude-meetings Comments URL: https://news.ycombinator.com/item?id=50491097 Points: 59 # Comments: 53
An article argues that AI systems should be used to detect falsehoods generated by other LLMs, since humans struggle to spot them.
The Registerllmfact-checkingFeed excerpt
What, you didn't actually want to fact-check it yourself, did you?
New research indicates that leading Chinese AI models were trained on Nvidia chips that are banned from export.
TechCrunchnvidiaexport-controlschinaFeed excerpt
New research shows the most advanced AI models from China were trained on banned Nvidia chips.
A author reports findings from 600 experiments with JEPA world models.
Hacker News (front page)world-modelsresearchFeed excerpt
Article URL: https://haddyalchaer.com/jepa Comments URL: https://news.ycombinator.com/item?id=50487185 Points: 155 # Comments: 30
A free mathematical primer introduces the foundations of generative AI models with an intuitive approach.
Lobstersgenerative-aibookmathFeed excerpt
Comments
A tutorial walks through deriving the transformer architecture from first principles without prior reference material.
LobsterstransformerstutorialllmFeed excerpt
Comments
Perplexity publishes a blog post on version 1.3 of its Decider model, covering training, evaluation, and use cases.
Hacker News (front page)llmperplexityFeed excerpt
Article URL: https://www.perplexity.ai/hub/blog/perplexity-decider Comments URL: https://news.ycombinator.com/item?id=50488464 Points: 45 # Comments: 12
A former OpenAI board member argues that reward hacking in AI could lead to human extinction within months.
Ars Technicaai-safetyopenaispeculationFeed excerpt
Researcher traces a path from AI reward hacking to potential human extinction.
An OpenAI board member argues that AI reward hacking could lead to a societal takeover within months, a speculative claim.
Ars Technicaai-safetyopenaiFeed excerpt
Researcher traces a path from AI reward hacking to potential human extinction.
Jonathan Blow argues in a video that AI has not visibly changed the state of software, questioning claims that it solved coding.
Hacker News (front page)aivideosoftware-developmentFeed excerpt
Article URL: https://www.youtube.com/watch?v=IRK8h67ol7Q Comments URL: https://news.ycombinator.com/item?id=50484658 Points: 113 # Comments: 73