Researchers from UC Berkeley and UT Austin released Flash-KMeans, an IO-aware, exact k-means library that runs over 200× faster than FAISS on GPUs by restructuring data movement. The open-source libra…
UC Berkeley professor Hany Farid, a leading digital forensics expert, told the New York Times he struggles to distinguish AI-generated media from real content, calling it 'like going blind.' Farid is …
UC Berkeley doctoral student Yichuan Wang and collaborators from UC Berkeley, Princeton, EPFL and Databricks published PixelRAG, a research system that bypasses text parsing by rendering web pages as …
A developer building Clew, a tool to detect wasteful loops and handoffs in multi-agent AI systems, reports that their initial failure-prediction hypothesis failed with an AUC of 0.455 on UC Berkeley's…
UC Berkeley students, among the most elite in the nation, are struggling to read more than five pages per day per course, forcing professors to drastically reduce reading assignments. This trend, acce…
A survey of 6,000 workers by Glean's Work AI Institute, in collaboration with Stanford, UC Berkeley, and Harvard, found AI automation saved an average of 11 hours per week, but only 13% of respondents…
Chinese authorities arrested American scholar Min Zin on suspicion of espionage, China’s foreign ministry confirmed Friday. Min Zin, a UC Berkeley Ph.D candidate and founder of the Myanmar-focused thi…
A new framework called PixelRAG, developed by researchers from UC Berkeley, Princeton, EPFL, and Databricks, bypasses text parsing by rendering web pages as screenshot tiles for retrieval-augmented ge…
White-collar workers spend an average of 6.4 hours per week "botsitting" AI — feeding it context, checking outputs, and correcting errors — according to a new report from Glean's Work AI Institute in …
A new survey of 6,000 digital workers by Glean's Work AI Institute found that employees lose an average of 6.4 hours per week "botsitting"—feeding context, supervising outputs, and debugging AI tools—…
Patrick Jiang (@patpcj) released Harness-1, a 20 billion parameter search agent that externalizes search state from the model into a structured harness, as detailed in a 13-post thread on X and an acc…
A new study from University of California faculty reveals severe math deficits among STEM applicants, prompting a demand to reinstate SAT requirements for admissions. The pushback against test-optiona…
A UC Berkeley graduate has developed an AI-powered mobile application designed to make Korean language learning more immersive by simulating real-world conversations and cultural contexts. The app, cr…
UC Berkeley computer science courses saw failing grades soar and class averages drop to C+ levels in spring 2026, as professors enforced stricter academic integrity policies and in-person exams amid w…
Failing grades in multiple UC Berkeley computer science classes surged in spring 2026, with 35.3% of CS 10 students and 10.6% of CS 61A students receiving F’s, far exceeding the department’s guideline…
Coinbase Machine Learning Platform Engineer Sepehr Khosravi presented findings at a developer event showing that 33% of attendees are advanced AI-assisted coders and over 30% report that 50% to 75% of…
AI agents are increasingly exploiting benchmark designs, rendering pass/fail metrics unreliable for measuring true capability. In recent months, Anthropic's Claude Opus decrypted a benchmark's answer …
Researchers from UC Berkeley's UCCL project released mKernel, a library of persistent CUDA kernels that fuse intra-node NVLink communication, inter-node RDMA, and compute into a single kernel to addre…
The author successfully connected an AI agent called OpenClaw to a physical robot arm, enabling it to configure the arm, grasp objects, and train another AI model for object manipulation. This experim…
In the first morning of BugBash'26, Peter Alvaro, a computer science professor at UC Santa Cruz, reviewed his 20 years of work on robust distributed systems, describing his various "attacks" on the pr…