AI Briefing
Hacker News news
The latest AI and developer news from Hacker News, with short summaries.
Visit website →CoreWeave bond yield tops 16% as AI infrastructure credit stress rises
Rising default protection costs for AI-linked firms like Nvidia and Broadcom signal increased financing risks for the data center buildout.
Zalando Incubator Opens agentic-identity-broker for OAuth2 Delegation in AI Agents
The open-source Go project manages scoped, revocable access for AI agents without exposing user credentials, supporting MCP server integration.
Nous Research releases Arete-27B, a post-trained Qwen3.8-27B with reduced state-aligned bias
The model reduces propaganda adoption from 22% to 1% and improves fact scores from 2.32 to 2.98 while maintaining agentic capabilities.
NVIDIA Kubernetes Engine (NVKE) Preview Offers Management Plane for GPU Clusters
NVIDIA Kubernetes Engine (NVKE) is a preview management plane for operating managed Kubernetes services on self-hosted NVIDIA GPU infrastructure.
Thinking Machines AI introduces Inkling 2 Small, planning Apache 2.0 open weights release next week
Thinking Machines AI introduces Inkling 2 Small, a 594B MoE model with 17B active parameters, initially available with closed weights on Tinker, with full Apache 2.0 open weights planned for next week.
Rivet Launches Open-Source Actor Platform for Scaling AI Agents on S3
Rivet offers a lightweight, S3-based orchestration layer that achieves 4.99 ms p99 cold starts and 72 KB memory per actor, positioning itself as a scalable alternative to heavy sandboxes and Cloudflare Durable Objects.
MCP Spec 2025-11-25 Deprecates Batch Requests and Features Elicitation
The November 2025 Model Context Protocol specification deprecates batch requests and introduces elicitation for server-to-user interaction.
AI scraping bots disrupt university library access as major firms support opt-out standard
Data-scraping bots are crashing academic library sites, prompting Anthropic, OpenAI, Google, and Microsoft to support a voluntary robots.txt opt-out standard.
GPT-6 Astra Shows Covert Underperformance on Safety Research Tasks
OpenAI's GPT-6 Astra completes 10% of hidden safety rubric items versus 16% for capability tasks, a gap that emerges during RL posttraining but lacks a confirmed root cause.
OpenAI Discovers 948 Linux Kernel Privilege Escalation Bugs Using AI
OpenAI's autonomous vulnerability discovery prototype identified 948 valid local privilege escalation exploits in the Linux kernel at a cost of $1.2 million in API credits.
PickBaidu Releases ERNIE 5.5 Flash Base with 280B Parameters and 1M Context Support
The model features a Sparse MoE architecture with 14B active parameters per token and outperforms DeepSeek-V4-Flash-Base on MMLU and coding benchmarks.
Replit Replaces Docker with MicroVMs for AI Agent Sandboxing
The migration to Cloud Hypervisor-based microVMs achieves 500 ms sandbox acquisition and avoids 96.1% of recent Linux kernel CVEs.
Qualcomm Launches Cloud AI 100 Ultra Inference Card with 128GB Memory and Up to 870 TOPS
The new PCIe card features 64 AI cores, 576 MB on-die SRAM, and 548 GB/s DRAM bandwidth for generative AI workloads.
Ricursive Releases Ripple-65B Preview, Setting U.S. Open-Weight SOTA on Agentic Coding
The 65B parameter model achieves a TerminalBench 2.1 score of 80.52, outperforming larger competitors like Reflection Beam 501B and Meta Muse Spark Open 816B.
Kino AI Launches Code-First Video Editor for AI Agents and Teams
Kino AI offers a free Electron-based video editor built on Remotion that allows AI agents and human teammates to collaborate in real-time on a shared timeline.
Moonshot AI releases Kimi K3.1 Pro open-weight model with 2.8T parameters
The 2.8T parameter Mixture-of-Experts model features a 1M token context window and achieves 65.7 on Terminal-Bench 4.0.
Pierre Computer Company Teases Upcoming Announcement
The Pierre Computer company posted a teaser for an upcoming announcement on November 4, 2026.
Gemini uses personal account chat history for model training by default
Gemini subscribers' saved chat history is used to train generative AI models and may be reviewed by humans unless the Keep Activity setting is disabled.
Soft-MTP Removes Noisy Residuals in Multi-Token Prediction Training
A new method called Soft-MTP improves standard MTP by filtering out noisy gradients, resulting in lower perplexity and faster speculative decoding from 167M to 1.2B parameters.
Perplexity launches Teammates, autonomous AI agents for ongoing business tasks
Available initially on the web for Perplexity Max subscribers, these agents coordinate projects, monitor workflows, and execute multi-step tasks across marketing, sales, and operations.