New Technology– category –
-
CO2-to-Fuel Conversion: AI-Driven Catalyst Mapping at IISc
Indian Institute of Science (IISc) / ScienceDaily India Overview Researchers at the Indian Institute of Science (IISc) used machine learning to identify 9,389 chemical reactions on a copper catalyst, expanding a previous model of 152 rea... -
Modular Hyperbolic Codes: Google’s 30x QEC Overhead Reduction
arXiv (via Medium) USA Overview Google Quantum AI, Google DeepMind, and MIT have developed a modular quantum error correction (QEC) architecture that reduces physical qubit overhead by over 30x compared to standard 2D surface codes. By n... -
MIT AI Weather Model: Subseasonal Forecasting Accuracy for 2–6 Weeks
Tech Xplore / Phys.org (via ScienceDaily) USA Overview MIT researchers have developed a machine-learning "wrapper" that doubles forecast accuracy in the two-to-six-week subseasonal range. By correcting systematic errors in existing dynam... -
Wanma Technology: Liquid cooling specs for AI data center PUE
Wanma Technology Co., Ltd. China Overview Wanma Technology has introduced a line of liquid cooling products for hyperscale AI and High-Performance Computing (HPC) data centers. These products improve Power Usage Effectiveness (PUE) by 15... -
SWE-chat analysis: AI coding agent survival rates and security
arXiv Unknown Overview An arXiv paper, "SWE-chat," presents the first comprehensive dataset and empirical analysis of AI coding agent usage and failure modes in real-world settings. Key findings include a 59% survival rate for agent-gene... -
Kolibri 1 explained: Aleph Alpha’s open-weight MoE LLM specs
ActuIA Germany Overview German AI firm Aleph Alpha launched 'Kolibri 1,' an open-weight Mixture-of-Experts (MoE) large language model, under the Apache 2.0 license on October 3, 2026. This 78.1 billion parameter model activates only 3.46... -
Tensor Machines: Open-source benchmark for AI compute costs
PR Newswire USA Overview Tensor Machines has released an open-source AI hardware benchmark designed to measure the true cost of useful AI output. This benchmark links GPU performance and power consumption directly to tangible AI output, ... -
Reflection AI Beam: Specs for 501B model and 4x efficiency
NeoTiler USA Overview Reflection AI has announced "Beam," a 501-billion-parameter open-weight model designed for inference, coding, and agent tasks. Featuring a sparse mixture-of-experts architecture with 23 billion active parameters and... -
Danaher: AI autonomous lab at Abcam speeds discovery by 8x
Danaher Corporation USA Overview Danaher has launched its first AI-powered autonomous lab at Abcam, aiming to accelerate life sciences research, particularly in the discovery of affinity reagents like antibodies. This lab utilizes AI to ... -
LLM Reasoning: ‘Always-Last’ strategy vs self-consistency
arXiv USA Overview This paper investigates 'expansion utility,' the change in correctness from restarting a Large Language Model (LLM) reasoning chain at a stored step. Measuring this across nine models and six benchmarks, the study foun...
