UC Berkeley's RDI just launched Agents' Last Exam (ALE), a demanding benchmark testing AI on real desktop workflows across 55 industries. OpenAI's GPT-5.5 topped the board with a 24.0% pass rate, narrowly beating Anthropic's new Claude Fable 5 at 22.0%. Why it matters: even the best models scored 0.0% on the hardest tasks, proving AI agents still struggle with genuine professional workloads.
Researchers from UC Berkeley, Stanford, and Databricks just introduced GEPA, a prompt optimization method that ditches reinforcement learning. Instead of thousands of trial-and-error runs guided by simple scores, GEPA has models reflect on detailed execution logs and error traces, beating traditional methods with up to 35 times fewer trials.
Researchers at UC Berkeley and UCSF just developed an AI neuroprosthesis that turns brain signals into streaming speech using a patient's own pre-injury voice. The streaming model decodes motor cortex activity into audible words in under a second, down from eight seconds in 2023.
UC Berkeley’s NovaSky team just open-sourced Sky-T1-32B-Preview, a reasoning model that rivals OpenAI's o1-preview while costing under $450 to train. Fine-tuning Qwen-2.5-32B on 17K curated reasoning samples took 19 hours on 8 H100 GPUs. Best of all, the entire pipeline, data, and weights are publicly available.
