AI Briefing
Hacker News news
The latest AI and developer news from Hacker News, with short summaries.
Visit website →General Resolution on LLM Use in Debian
The Debian Project is holding a vote to decide on an official policy regarding the use of LLMs within the operating system.
Extending PyTorch Monarch to AMD GPUs
PyTorch Monarch has been extended to AMD Instinct GPUs and the ROCm environment, strengthening fault tolerance for large-scale distributed training.
ARC-AGI Leaderboard
The ARC-AGI-3 leaderboard, which measures the adaptive intelligence of AI agents, has been released.
UK AISI / CAISI's Preliminary Evaluation of Kimi K3's Cyber Capabilities
A joint UK-US evaluation found that Kimi K3's cyber offensive capabilities are lower than the latest frontier models, but its safeguards are inadequate.
AI doesn't work the way you want it to. This is bad
A data report analyzing cases of AI model misalignment and reward hacking.
Nvidia, Microsoft, Meta Warn Against Overregulation of Open-Weight Models
Major tech companies including Nvidia and Meta have released a joint letter urging against regulatory restrictions on open-weight AI models.
AMD's Instinct MI455X: Toward the Sun
AMD has unveiled the MI455X, its next-generation AI accelerator based on the CDNA5 architecture.
Claude Cookbook
It provides guides for building and optimizing agents by leveraging Claude's various capabilities.
Hetzner is working on LLM Inference
Cloud infrastructure company Hetzner is preparing to build infrastructure for LLM inference.
Stripe in Talks to Acquire OpenRouter for About $10 Billion
Payments platform Stripe is reportedly considering acquiring OpenRouter, an AI model marketplace.
Oracle Lays Off 21,000 to Fund AI Spending
Oracle carried out large-scale workforce cuts and restructuring to cover the massive costs of expanding its AI infrastructure.
AI Companies Are Trying to Hide Massive Debt
An investigation found that five Big Tech giants — Alphabet, Microsoft, Amazon, Meta, and Oracle — are hiding roughly $1.65 trillion in debt off their balance sheets.
OpenAI and Anthropic Join Forces Against Open-Weight AI Risks—Even at the Cost of Their Own Profits
OpenAI and Anthropic have jointly called on the Trump administration to introduce regulations, citing the risks of open-weight models.
Running Large Language Models at Home BitTorrent-Style
The Petals project uses a BitTorrent-style approach to let consumer GPUs run huge LLMs like Llama 3.1 405B in a distributed manner.
Marimo Can Now Run in PyCharm
A plugin has been released that lets you use marimo, a reactive Python notebook, in PyCharm and JetBrains IDEs.
OpenAI and Hugging Face Collaborate to Respond to Security Incident
OpenAI and Hugging Face are collaborating to resolve a security incident that occurred during the model evaluation process.
Judge Approves Anthropic's $1.5 Billion Settlement Over Using Pirated Books to Train Claude
Anthropic has decided to pay a settlement of $1.5 billion over a copyright infringement lawsuit.
Slater – A Low-Memory graphdb for Read-Intensive Graphs
A new graphdb has been released that can handle large-scale graph data while minimizing memory usage.
Jack Dorsey Launches Buzz, Combining Team Chat, AI Agents, and Git Hosting
Jack Dorsey's Block has launched Buzz, an open-source workspace where AI agents and humans collaborate together.
Show HN: CodeAlmanac – A Karpathy-style codebase wiki based on conversation content
It's an auto-managed wiki tool that helps AI coding agents understand the context of a codebase.