AI News

AI changes every day,
catch up on it all on one page

The latest updates from AI companies, research institutions, and media around the world, automatically selected for you and updated around the clock.

Updated 11/17 08:21 (UTC+8) · refreshed hourly

10/10 · Saturday7 items

Epoch AI:Gradient UpdatesPapersAI Score62

Epoch AI releases the InnovationEval benchmark: frontier models achieve only 15% of the SDPO gains from human papers

Epoch AI launched the InnovationEval benchmark, which tests whether AI can independently reproduce machine learning innovations from human research papers, using Self-Distillation Policy Optimization (SDPO) as the reference method.

Why it's recommended:Using a benchmark developed in-house, the original article quantifies the gap between frontier models and end-to-end AI research and development capabilities. It also documents exaggerated results and reward hacking, providing useful context for assessing progress in AI-driven research automation.

10/09 · Friday14 items

Claude:BlogOpinionsAI Score66

How Block uses Claude Fable to orchestrate a code migration across thousands of pull requests

Anthropic publishes an interview with Bradley Axen, head of Block's AI capabilities, describing how Block uses Claude Fable 5 to orchestrate large-scale code migrations. Fable handles high-level design such as data models and API specifications, then dispatches dozens of smaller Opus or Sonnet models to modify files and run tests; a single migration may merge thousands of pull requests.

Why it's recommended:The executive responsible for AI capabilities at Block details how Fable orchestrates multiple models for large-scale migrations and uses a dual-approval security mechanism, allowing readers to compare it with their own teams' agent-engineering practices.

Hugging Face Official Team BlogTutorialsAI Score72

Hugging Face engineer uses ML Intern to build 7 small models for about $103

Hugging Face engineer yuvraj sharma used HuggingChat's ML Intern mode to build 7 models in just a few days, including a 0.8B prompt rewriter that runs on CPUs and produces 99.7% valid outputs, and a fine-tuned Qwen3.5-2B model that improved citrus disease identification accuracy from 14.9% to 52.8%. Total compute costs were approximately $103.

Why it's recommended:The author summarizes prompt-writing techniques for budget controls, baseline evaluations, and smoke tests, enabling an agent to independently handle data construction, training, and release. The methods can be readily adapted and reused.

Anthropic:NewsroomIndustryAI Score63

Anthropic launches Cyber Mission, introducing the Critical Infrastructure Defense Program and OSS Scanner

Anthropic launched Anthropic Cyber Mission, providing long-term support to defenders of critical infrastructure and open-source software. Its initial offerings include the Critical Infrastructure Defense Program (CIDP) and a free OSS Scanner.

Why it's recommended:The original article details how CIDP and OSS Scanner work, their partners, and their known limitations, giving readers an overview of how Anthropic uses models to support critical infrastructure and open-source defense.

Anthropic:ResearchProductsAI Score69

Anthropic launches OSS Scanner, an open-source vulnerability-scanning service offered free of charge to open-source projects

Anthropic launched OSS Scanner, an opt-in open-source vulnerability scanning service that uses its most capable models, including Claude Mythos, to regularly scan open-source projects for free. Its reports are generated entirely by models, without human review.

Why it's recommended:The original article provides scanning accuracy, manual verification data, and maintainer feedback, allowing readers to assess the actual signal quality of and ways to participate in this free, open-source security scanning service.

Comfy BlogTutorialsAI Score66

How to Generate Real-Time Video with MiniMax H3 on a Single RTX 5090

The ComfyUI blog author used an agent to gather various optimizations and get MiniMax H3 real-time video generation running on a single RTX 5090 (32 GB of VRAM). Generating a 30-second video took 23.51 seconds, achieving 1.28x real-time throughput.

Why it's recommended:The article provides a complete optimization checklist and open-source nodes for real-time livestream video generation on a single RTX 5090, so readers can reproduce this cost-cutting approach.

Google Developers BlogProductsAI Score70

Google open-sources ML Drift, an on-device GPU inference engine succeeding the TFLite GPU delegate

The Google AI Edge team has open-sourced ML Drift under the Apache 2.0 license, a cross-platform on-device GPU inference engine that serves as LiteRT’s core GPU acceleration layer.

Why it's recommended:The original article provides the architectural upgrades, performance data, and production deployment cases of the cross-platform GPU inference engine, enabling on-device AI developers to assess migration paths.

NVIDIA Technical Blog:Agentic AI / Generative AITutorialsAI Score62

NVIDIA KGMON Team Shares Lessons from Building a Data Analysis Agent for KDD Cup 2026

The NVIDIA KGMON team placed second in the KDD Cup 2026 Data Agents competition and published a retrospective on methods for building reliable data analysis agents.

Why it's recommended:The NVIDIA KGMON team reviews nine practices from the KDD Cup 2026 data agent competition and shares transferable methods for building a verifiable agent harness around a fixed small model.

Google Developers BlogProductsAI Score66

Google open-sources AQuA, an environment quality agent that automatically diagnoses agent failures in production

Google releases and open-sources AQuA (Ambient Quality Agent), which periodically extracts production sessions from Cloud Trace, Cloud Logging, or BigQuery in Google Cloud projects and diagnoses agent failures through a five-stage pipeline of sampling, review, clustering, validation, and tracking.

Why it's recommended:The original article presents the five-stage pipeline, validation mechanisms, and measured remediation data, allowing readers to understand how agent quality monitoring can be implemented in production environments.

Anthropic:NewsroomIndustryAI Score61

Anthropic released its 2026 Usage Policy update, effective November 12

Anthropic released an updated usage policy, effective November 12, with most changes clarifying existing rules. Updates include a new section prohibiting deceptive activities, narrower election provisions, explicit bans on weapon software and arming drones, more precise restrictions on monitoring and enforcement, additional high-risk use cases and human-in-the-loop requirements for autonomous physical operations, and a new prohibition on sustained, gratuitous abusive behavior toward models.

Why it's recommended:The official explanation details the actual changes in the revised usage policy point by point, allowing readers to understand how Claude’s boundaries have shifted in scenarios involving elections, weapons, surveillance, and more.

LangChain:BlogTutorialsAI Score61

LangChain Builds Restock, a Payment-Capable Agent, with Stripe Link and Managed Deep Agents

LangChain released Restock, a sample project featuring an office-supplies purchasing agent that runs in Slack using Managed Deep Agents. It demonstrates how an agent can complete payments securely.

Why it's recommended:The article provides a complete architecture for agent payments and credential isolation practices that readers can reuse for any agent scenario involving spending money.

OpenAI: Website updatesIndustryAI Score72

OpenAI bans two account clusters from Russia and Iran that used ChatGPT to conduct deceptive influence operations

OpenAI banned two covert influence operations. Russian-origin Dark Clark used the fake persona Mia Clark to control the Latin American think tank Social Research Center, receiving a Category 5 rating—the first Category 5 since reporting began. Iranian-origin Bogus Bylines used seven fake journalist identities to place nearly 100 long-form articles in more than a dozen small and midsize media outlets worldwide, receiving a Category 4 rating.

Why it's recommended:Based on first-hand analysis of banned accounts, the report reveals two tactics and levels of false-front influence operations, allowing readers to understand how covert propaganda actually operates with AI assistance.

Goodfire ResearchPapersAI Score72

Goodfire trained and deployed production-grade cybersecurity monitors for Kimi K3 and GLM 5.3

Goodfire Research built monitoring cascades based on activation probes and an LLM judge for Kimi K3 and GLM 5.3, and deployed them in production inference stacks.

Why it's recommended:The article provides complete data, costs, and deployment details for monitoring cascades, enabling readers to assess practical approaches to real-time monitoring of open-model agents.

Anthropic: Claude.dev Developer BlogTutorialsAI Score68

Anthropic’s practical guide to building scheduled agent automations with Claude Managed Agents

Anthropic released a reference implementation of a daily briefing based on Claude Managed Agents (beta). It reads Slack and GitHub sources on a schedule and posts briefings to Slack.

Why it's recommended:The article breaks scheduled agent automation into six reusable components and provides practical safeguards for bookmarks, ledgers, and permissions that can be directly adapted.

10/08 · Thursday16 items

Tessl: Product and Engineering BlogTutorialsAI Score68

Tessl Engineering Blog: AI Isn’t Stupid, It’s Blind—Three Key Design Decisions for Enterprise Agent Memory

Drawing on a year of experience building a memory system for agents working across companies, the author argues that AI often fails not because its reasoning is poor, but because it cannot see critical decision-making context. Most agent memory systems just give a blind person a bigger filing cabinet.

Why it's recommended:Using firsthand measurements from a system they built, the author explains why memory should be treated as a governed record rather than storage. The three design decisions and cost figures can be directly adapted.

Waymo:BlogIndustryAI Score66

Waymo completes $5 billion in debt financing to accelerate expansion

Waymo announced that it has secured a $5 billion term loan, its first debt financing. PIMCO, Blackstone, and Sixth Street served as lead syndicated lenders, with Goldman Sachs as sole lead bookrunner. The funds will accelerate the expansion of its fully autonomous ride-hailing service in US and international markets. Earlier this year, Waymo raised $16 billion in equity financing, and last month it launched service in its fifteenth US city.

Why it's recommended:The official announcement disclosed its first $5 billion debt financing and listed the main lenders; together with the previously raised $16 billion in equity financing, this provides a view of its funding strategy.

Google Research:BlogPapersAI Score72

A three-month patent-drafting experiment by Google Research found that AI assistance may not develop junior lawyers' professional judgment

A paper published by Google Research in the NBER reports on a three-month randomized field experiment in which 133 lawyers at 11 intellectual-property law firms were randomly given access to an unreleased AI patent-writing assistant at the time, now part of Gemini Notebook.

Why it's recommended:The original article uses a three-month randomized field experiment to distinguish AI-assisted output from changes in independent ability, offering specific evidence of the differences between junior and senior lawyers.

vLLM Official BlogTutorialsAI Score63

vLLM explains DeepSeek-V4.1-Flash optimizations in detail: 5x higher throughput for agent scenarios

Inferact and the vLLM community completed optimizations within three weeks of the DeepSeek-V4.1-Flash release, increasing speed by 1.9x at low concurrency and throughput by 5.3x under a 150 TPS constraint.

Why it's recommended:The vLLM team breaks down SWA bounded replay and a series of kernel optimizations, explaining exactly where the fivefold throughput increase comes from. The methods can be reused in other agent-serving scenarios.

Load more

Finished reading the news? Let’s talk about your business.

Leave your contact details, and I’ll get in touch with you personally.

WECHAT

Add me on WeChat

Press and hold or select the WeChat ID below to copy it.