Research1 min read
AI Costs Drop 900x; New Data Systems Needed for Agents
Inference costs fell between 9 and 900 times per year. Researchers propose redesigning data systems for swarms of agents.
From Berkeley AI Research
Loading
Blog
Daily notes on new models, LLM releases, agent frameworks and AI research, written from the sources we follow and delivered as a newsletter every day.
Get the morning and evening issues
Every new post of the day, in one email. Confirmation required.
Research1 min read
Inference costs fell between 9 and 900 times per year. Researchers propose redesigning data systems for swarms of agents.
From Berkeley AI Research
Research1 min read
The Berkeley AI Research Lab celebrated its class of 2026 Ph.D. graduates on July 1, 2026. These researchers work on robotics, large language models, and AI safety.
From Berkeley AI Research
How this blog is made
POST /api/v1/route with execute: true.Photo: Yan Krukau
Machine Intelligence
From Google Research blog
LLMs1 min read
Privacy controls — systems that enforce retention, access, allowed-purpose, downstream-sharing, or anonymization policies — require a reliable understanding of data to function. Before such a control can operate effectively, it must know...
From Meta AI engineering
AI1 min read
What if your cloud environment could help you move from insight to action in real time, with systems already working through the next set of decisions? The post From insight to action: The next phase of agentic cloud operations appeared ...
From Microsoft Azure AI blog
Research1 min read
Google DeepMind is implementing an AI Control Roadmap to secure internal systems, combining traditional safeguards with real-time monitoring of AI agents. This approach focuses on operational resilience and ongoing safety measures for deployed agents.
From Google DeepMind blog
AI1 min read
At Microsoft Build 2026, AI moved from experimentation to execution—shifting from isolated tools to connected systems grounded in business data. Organizations that win will embed AI across workflows, scale it effectively, and translate i...
From Microsoft Azure AI blog
Research1 min read
Amazon Science released four methods to stop AI agents from breaking physical laws. These techniques improve accuracy and safety for robots and systems that operate in factories and hospitals.
From Amazon Science
Research1 min read
Amazon scientists launched a new open-source harness to fix the gap between what AI models intend and what tools actually do. The system improves reliability across multiple benchmarks without needing specific tuning for each model.
From Amazon Science
Models1 min read
Mistral launched Emmi AI to speed up engineering design and manufacturing.
From Mistral AI news
LLMs1 min read
We’re introducing SilverTorch, a reimagining of recommendation systems that unifies all retrieval components for user generated content under a unified architecture. SilverTorch shows up to 23.7x higher throughput compared to the state-o...
From Meta AI engineering
Posts are drafted from public feeds by models OpenSmartRoute routes to - the same router, skill and metering customers use - and always link to the original source. Corrections: support.
Archive (93)