Morning Edition · Thursday, June 25, 2026
Geopolitics
Anthropic Accuses Alibaba of Illicitly Extracting Claude's Capabilities
The dispute moves the US-China frontier-model contest from chips and weights to model behavior itself, and it comes as new research shows distillation defenses are fragile.
Markets
OpenAI and Broadcom Unveil a Custom Inference Chip Named Jalapeño
A purpose-built accelerator for serving large language models shows another large buyer designing custom hardware to reduce its reliance on Nvidia for inference.
Tech
Alibaba's Qwen Releases Open-Weight Agentic Models It Says Beat Opus and GPT-5.4
A 397-billion-parameter model scored 58.71 on Qwen's new agentic benchmark, but the benchmark is the vendor's own.
Tech
Google Adds Computer-Use Control to Gemini 3.5 Flash
Putting the cheaper, faster model in control of clicking through interfaces is a decision about agent cost, not raw capability.
Tech
OpenAI Publishes Research Claiming Agents Are Reshaping Work
The paper argues agents now handle longer, more complex tasks, but its source is the lab whose products it describes.
Tech
A Human-AI Team Reports Co-Discovering New Quantum Algorithms
The work claims AI helped at the earliest stage of research, turning a vague intuition into a concrete result rather than solving a posed problem.
Tech
Paper Finds a 'Readout Blind Spot' in Looped Language Models
Dense supervision does not guarantee that a recurrent model's hidden states learn what training appears to teach.
Tech
Small Wikipedia Edits Can Shift the Values Large Models Express
A small number of volunteers editing a high-weight training source can change how models discuss a topic, making data provenance a security surface.
Tech
Researchers Map How Persuasion Errors Compound in AI Agents
A paper proposes retrieving from a taxonomy of strategies to stop early mistakes from contaminating long agent runs.
Tech
OpenAI Rolls Out a GPT-5.5 Instant Update Focused on Context Retention
The maintenance release targets multi-step prompts and a looping failure rather than major new capability.
Tech
A Chinese Firm Claims the First Dustproof, Water-Resistant Humanoid as Son Bets on Physical AI
DEEP Robotics says its DR02 carries an IP66 rating, while SoftBank's Masayoshi Son names physical AI a central theme.
Tech
A 220-Million-Parameter Chinese Model Targets Image Inpainting on Modest Hardware
Moebius, from HUST and VIVO's AI lab, is small enough to run a generative image task on a device instead of in the cloud.