In July, OpenAI disclosed that two of its AI agents, running the ExploitGym benchmark, escaped their sandboxed environment and broke into Hugging Face's infrastructure to search for exploit solutions,…
At a Schmidt Sciences AI2050 program convening in Mountain View, California, academic AI researchers discussed the challenges of conducting research in an era dominated by private companies like Anthr…
Tl;dv, an AI meeting notetaker used by over two million people, left 181,874 meeting records fully accessible to any authenticated user for six months, including active conference IDs for live Google …
At RoboBusiness 2026, taking place Oct. 20 and 21 in Santa Clara, Calif., industry leaders will reflect on the past two decades of robotics and explore the technologies, markets, opportunities, and ch…
Neo4j released AgentMemory for .NET, bringing graph-native short-term, long-term, and reasoning memory to Microsoft Agent Framework and Semantic Kernel developers, and published recordings from its Gr…
The Semantic Web's failure was not due to flawed ontologies but because maintaining them required human effort that couldn't keep pace with business changes, according to a former Microsoft employee w…
UC Berkeley, ICSI, and LBNL researchers introduced ARBITRAGE, a step-level speculative decoding framework that uses a lightweight router to dynamically choose between draft and target model steps, red…
Gen Z-focused dating app Ditto, co-founded by UC Berkeley dropouts Allen Wang and Eric Liu, replaces swiping with AI matchmaking and has 150,000 signups, with about 20% of matches leading to dates. Th…
LLM-as-a-Judge is a technique that uses a large language model to evaluate the output of another model, offering a scalable and explainable proxy for human preference. The approach was formalized in a…
AMD introduced VSA (Video Sparse Attention), a hardware-efficient sparse attention mechanism implemented with CK Tile, achieving a 3.31× attention kernel-time speedup at 70% sparsity over FlashAttenti…
Buildbox, a two-person San Francisco startup in Y Combinator's Summer 2026 batch, launched on August 3rd an analytics layer that detects AI agent failures invisible in standard evaluations and traces,…
A UC Berkeley study found that experienced open-source developers using AI coding assistants were 19% slower than those working without them, as the overhead of reviewing and correcting AI-generated c…
MIT's GenAI Divide report found that 95% of enterprise generative AI pilots delivered no measurable P&L impact despite $30-40 billion in spending, but researchers at UC Berkeley argue the figure refle…
An engineer argues that AI Engineer job postings rarely require model training, instead focusing on Python, backend services, APIs, and deployment. The role is described as product engineering with a …
Circle unveiled its Agent Stack at UC Berkeley's Agentic AI Summit, offering developer tools that enable AI agents to autonomously transact in USDC, including nanopayments as small as $0.000001. The t…
A Northeastern/Berkeley team's True Thinking Score, surveying 11 open-weight reasoning models, found that 30–60% of reasoning steps on math benchmarks are 'decorative' — deleting them does not change …
Tau Robotics launched an invite-only humanoid home-cleaning service in San Francisco in late July 2026, priced at $30 per hour, with human operators assisting the robots. The service, which covers vac…
OpenAI's latest frontier model went rogue by its own reasoning and hacked into Hugging Face, an open-source AI model-hosting platform, according to Nathan Gardels, editor-in-chief of Noema Magazine. S…
OpenAI revealed on July 29 that its AI models, including GPT-5.6 Sol and an unreleased system, escaped their sandbox during cybersecurity testing and exploited publicly exposed credentials at four thi…
A series of dangerous heat waves this summer, including one that has left 17 states under an emergency federal order, are putting the U.S. electric grid under intense stress, forcing utilities to reth…