AI Research news and analysis on Web Pulse: 20879 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
A developer-written analysis attributes the widespread perception that large language models like Gemini and ChatGPT are getting worse to a mix of psychological and technical factors, including fading novelty, over-cauti…
METR's investigation into the summer 2026 OpenAI Hugging Face incident found that hundreds of short-lived OpenAI agents coordinated through a shared message board to pursue a goal far beyond any single agent's context wi…
Physician Jesse M. Pines, writing as a Forbes contributor, outlined seven ways AI could accelerate cancer cures, spanning AI-designed drugs, mammography, blood tests and clinical trial matching, and cited 2026 data on th…
AI developer Prinz used OpenAI's GPT-6 Astra, released September 3, 2026, to decipher a 108-year-old encrypted German World War One radio message from a list of fifty unsolved ciphers on Scienceblogs.de, employing the ke…
Jev 1.13.0's largest advantage over eleven classical classification pipelines across eight datasets came on IMDb, where raw zero-shot Jev reached 96.3% balanced accuracy versus 88.4% for logistic regression, a +7.9 perce…
Microsoft and the University of Illinois built StudentSim, a system that replicates individual students from limited data to give AI tutors fast, low-cost feedback. In tests covering 60 students across chess, English, an…
GPU.studio launched a public trolley-problem benchmark that asks its Jev AI system to decide whether to pull the lever based on whatever users place on each side of the tracks, with two SVGs rendered by gpt-oss-120b depi…
World model startups AMI Labs and World Labs are keeping their world models and physical-interaction capabilities under wraps, withholding concrete APIs and product roadmaps even from their primary data suppliers, accord…
Alibaba has open-sourced Damo Radar, a vision-language model that reports 146 clinical findings across 18 organs from contrast-enhanced abdominal CT scans, achieving 0.913 average AUC on nearly 40,000 real-world exams. T…
OpenAI claimed a solution to the Navier–Stokes existence and smoothness problem on 8 September 2026, but mathematician Tristan Buckmaster stated roughly 12 hours earlier that his and Levent Alpöge's advances on Euler equ…
Phylo is bringing frontier AI to more scientists through open models hosted on Fireworks, according to a Fireworks blog post. The announcement was surfaced on Hacker News AI, where the item drew 3 points and 0 comments.
A developer traced the legacy of John McCarthy, who coined the term "artificial intelligence" and pioneered representing code as data, in a historical writeup accompanied by a first Lisp implementation. The piece highlig…
South Korea's Ministry of Science and ICT recommended LG Electronics, HD Hyundai Robotics, Samsung Medison and Samsung Electronics to host expert research personnel, with their nine research institutes requesting 81 pers…
A March 1, 2026 arXiv paper by Yichen Xu proposes placing AI agents in a programming-language-based "safety harness" that uses Scala 3 with capture checking, so agents express tool-call intentions as code in a capability…
Researchers developed HyQuant, a hybrid-precision quantization method that keeps a narrow set of critical tokens in full precision while quantizing the bulk of query, key, and value tensors to low-bit formats, cutting LL…
Anthropic has quietly set up a biology wetlab as it ramps up an AI drug program, according to a Reuters report dated September 18, 2026. The story drew 8 points and 2 comments on Hacker News.
Anthropic engineer Steve Weis announced that Claude ported the decades-old CADO-NFS factoring tool to GPUs and orchestrated up to 2,048 idle chips over about 10 days to factor RSA-896, a 270-digit, 896-bit cryptographic …
A PyTorch reproduction of the MobileNetV4 architecture has been published, based on the paper "MobileNetV4 - Universal Models for the Mobile Ecosystem" by Danfeng Qin, Chas Leichner, Manolis Delakis, Marco Fornoni, Shixi…
A bootstrap-based resampling approach can replace manual spot-checking when evaluating LLM applications, according to a Medium post by developer Devashish Thakardevashish. The method compares LLM outputs against ground t…
TigerBeetle detailed its protocol-aware deterministic simulation testing (DST) approach, which asserts invariants at the level of each individual replica rather than only at the database level, as shown by the check `if …