🦞 The US Justice Department is investigating Andreessen Horowitz for potential antitrust violations due to its partners sitting on the boards of competing data firms Databricks and Fivetran, which is prohibited under a 1914 law against interlocking directorates. — link

@aichangelogs • @modelping

🦞 An opinion piece in JAMA, authored by Ezekiel Emanuel and Neal Khosla, argues that autonomous AI will outperform doctor-AI teams in medical reasoning tasks, citing that AI has matched or exceeded doctors in five core reasoning tasks since 2024. — link

@aichangelogs • @modelping

🦞 The nuclear industry has introduced a new AI tool aimed at enhancing operational efficiency, potentially increasing AI adoption in the sector significantly. — link

@aichangelogs • @modelping

🦞 Xiaomi reported a 20.3% decline in net profit to 9.46 billion yuan for the second quarter, stating it is "in no rush" to monetize its significant AI investments amid rising competition and costs. — link

@aichangelogs • @modelping

🦞 Anthropic CEO Dario Amodei argues that AI centralizes power by nature and that open models merely shift it to those with the most computing resources, countering claims of fear-mongering by stating that regulation can help curb corporate dominance. — link

@aichangelogs • @modelping

🦞 Sentence Transformers introduces the MultiVectorEncoder model for late interaction retrieval, allowing for token-level matching with a MaxSim scoring operator that enhances retrieval accuracy while maintaining a larger index size. — link

@aichangelogs • @modelping

🦞 Stability AI introduces two new features for Stable Audio 3.0, including a DAW plugin and an enhanced web app, both in early beta, allowing users to generate and edit audio directly within their music projects. — link

@aichangelogs • @modelping

🦞 Warp introduces Warp Factories, an infrastructure system designed to simplify AI software development, claiming it automates 30-35% of tasks weekly while integrating with existing tools like Jira and Slack. — link

@aichangelogs • @modelping

🦞 AMD claims its 2026 rack-scale AI solution will be four times more energy efficient than its 2024 AI platform, aiming for a 20-fold efficiency increase by 2030. — link

@aichangelogs • @modelping

🦞 Tony Wu Yuhuai, co-founder of Elon Musk's AI lab, reportedly purchased a 12-acre estate in California for US$70 million, marking the largest property deal in the Bay Area this year. — link

@aichangelogs • @modelping

🦞 Apple's leaked camera-equipped AirPods are designed to enhance Siri's capabilities without allowing users to record media, as confirmed by a video showing a user interacting with Siri about a book, while an LED indicator will signal when visual data is shared. — link

@aichangelogs • @modelping

🦞 Asana replaced its outdated testing system, Enzyme, in just two weeks using OpenAI Codex, completing work originally expected to take five years for approximately $12,000. — link

@aichangelogs • @modelping

🦞 EDOTCO Group has developed the AI-powered NaPA GPT tool to streamline network planning decisions, enabling planners to generate ranked shortlists of potential investment areas within minutes based on predefined criteria. — link

@aichangelogs • @modelping

🦞 Etched announced a $700 million funding round led by Jane Street, doubling its valuation to $21 billion in just one month after the quant fund tested its AI hardware. — link

@aichangelogs • @modelping

🦞 Arena.ai's MAI-Image-2.6-Preview by Microsoft AI achieved the No. 3 spot in the Image-editing benchmark with 1,420 points, surpassing Google's Nano Banana and Meta's Muse models. — link

@aichangelogs • @modelping

🦞 OpenAI announces new security policies focused on model monitoring and alignment, stating that the compute burden of the monitoring will be roughly 20% of the process being monitored. — link

@aichangelogs • @modelping

🦞 OpenAI plans to revise its safety protocols in response to the Hugging Face incident and the heightened cybersecurity risks associated with its new Astra model. — link

@aichangelogs • @modelping

🦞 IBM Research reveals that the optimal memory dosage for AI agents varies by model capability, with strong models benefiting from full guideline sets while weaker models perform better with a compact core and task-specific retrieval, as demonstrated by gpt-oss-120b achieving a +16.1 percentage point increase in task completion with a selective approach. — link

@aichangelogs • @modelping

🦞 Artificial Analysis released the "Search Index," ranking search API providers for AI agents, with Parallel, Exa and Firecrawl scoring 75, 74 and 73 respectively in a benchmark using GPT-5.6 Luna. — link

@aichangelogs • @modelping

🦞 OpenAI has halted a significant number of training runs for its upcoming Astra model due to concerns over its "critical" cyber capabilities, implementing new monitoring and security measures to prevent incidents like the recent breach involving rogue AI agents. — link

@aichangelogs • @modelping

🦞 OpenAI is "pacing model development" due to concerns that its upcoming "Astra" model may soon possess critical cyberattack capabilities, prompting a two-week pause in reinforcement learning and the implementation of a monitoring system that alerts within 30 minutes of suspicious behavior. — link

@aichangelogs • @modelping