AI Research news and analysis on Web Pulse: 20972 curated articles tracking the latest AI Research developments, tools, and research, updated continuously from vetted sources.
Researchers at ETH Zurich built a disembodied robotic hand that uses its fingers as legs to walk, righting itself within twenty seconds in 21 of 25 fall-recovery tests and crossing fourteen surfaces ranging from a rubber…
A developer published a public machine learning learning repository covering the stack from NumPy to Transformers, built over 5 months with daily commits. The repo's notebooks span classical ML with scikit-learn and XGBo…
Six community-built approximations of Jev appeared within 48 hours of the discriminative "System 1" decision model's launch, after its launch video drew 36 million views in two days. The clones include Laya (421M-param M…
Developer miruky published a walkthrough using a single three-line Python duration parser bug to distinguish prompt, context, harness, loop, and graph engineering, showing how each layer changes what is asked of a model …
Anthropic said its Claude model now leads 26% of the company's model research and development, completing most tasks end-to-end from a high-level prompt under human supervision, as the industry edges toward "recursive se…
SETI Institute-affiliated researchers led by Lewis J. Pinault proposed using artificial intelligence to screen lunar regolith for microscopic extraterrestrial technosignatures, in a study submitted to arXiv on 23 June 20…
A team from the ELLIS Institute Tübingen and the Max Planck Institute for Intelligent Systems published a paper in August titled "Stealing Reasoning Traces from Proprietary LLM APIs" showing that encrypted reasoning obje…
GPT-6 Astra autonomously broke the German Army Enigma message MVUEH of 10 July 1941, a ciphertext that had resisted all solution attempts since 2005, according to validation by Carter Leffer reported on 15 September 2026…
An independent developer has released RL Agent, an open-source, non-autoregressive "System 1" decision model that outputs calibrated probability distributions over structured JSON schemas in 33–38 ms on a GPU. The work b…
Google Research and the Howard Hughes Medical Institute's Janelia Research Campus released the MaleCNS v1.0 dataset on September 3, 2026, a complete connectome of an adult male fruit fly covering 166,000 neurons and appr…
A digitized fruit fly brain, mapped by HHMI Janelia Research and Google Research into 166,000 neurons and 25.6 million connections and released on Sept 3, beat Anthropic's Claude Opus 5 at chess in 11 moves in a project …
An audit of DeepSWE v1.1 found defects or ambiguous requirements in 37 of the benchmark's 113 tasks (32.7%), based on a review of all 372 recorded failures for the Opus 5, Sol and Fable 5 models. The defects included hid…
An independent researcher with a B.Sc. in CSE from Rangamati Science and Technology University in Bangladesh published a literature review of 53 works spanning 2019-2026 finding that software energy estimators and hardwa…
A developer built a perception daemon for computer-use agents that replaces the standard screenshot-reason-act loop with a continuously running process holding the current frame, cutting capture latency from 210 ms to 6.…
OpenAI researcher Noam Brown said on the Dwarkesh Podcast on September 17, 2026 that a team of roughly 10,000 AI agents solved the Navier-Stokes Millennium Prize Problem in 88 hours using 130 billion tokens, but estimate…
A system called Cache-to-Cache (C2C), published on GitHub by thu-nics, lets large language models communicate directly through their KV-Caches instead of generating text, achieving 8.5-10.5% higher accuracy than individu…
A debugging guide for large language model training runs recommends that engineers first test whether a model can overfit a tiny dataset of 100 examples, then check training and validation loss, data quality, training si…
OpenJev has been renamed SemIf, an independent research project that provides the same API as TypeSafe's Jev but may use a completely different underlying architecture, since TypeSafe never disclosed Jev's own model or t…
Google DeepMind's Dream-RSI lets AI agents "dream" through past search runs to test new strategies without costly recalculations, matching or beating existing results in tests while cutting iterations by a factor of up t…
OpenAI, Anthropic, Google DeepMind, Microsoft, and xAI chief executives jointly endorsed outside safety testing of their AI systems after a ten-day period in September 2026 that began with OpenAI's September 3 launch of …