Nous Research
Nous Research, the open-source AI lab behind the Hermes open-weight models, released Hermes Agent in February 2026 and the MIT-licensed agent had more than 260,000 GitHub stars as of 8 November 2026, a figure Nous cites …
Nous Research, the open-source AI lab behind the Hermes open-weight models, released Hermes Agent in February 2026 and the MIT-licensed agent had more than 260,000 GitHub stars as of 8 November 2026, a figure Nous cites …
Ai2, the Allen Institute for AI, publishes fully open model families including Olmo 3 in 7B and 32B sizes (Base, Think and Instruct variants) and Molmo 2 vision-language models in 4B and 8B sizes, all under the Apache 2.…
A proposal recommends adding a dedicated function parser page to the Fortran playground built on fparse, a Fortran function parser that supports expression systems, vectors, automatic derivatives and standalone Fortran c…
OpenAI published Agents SDK v0.24.0-alpha.7 as a prerelease on 2026-11-08, changing defaults across three configuration surfaces: new Environments now default to write_policy="strict" (requiring a compatible Linux or mac…
Element Labs, Inc.'s LM Studio desktop app now ships Bionic, an agent for open models launched in July 2026 that edits documents, writes code, runs automations and transcribes voice locally, and can switch to US-hosted o…
Tencent announced Hy4 on 4 November 2026, its latest large language model and the first in the Hy line with visual understanding and computer-use abilities, and named Tencent Yuanbao alongside WorkBuddy, CodeBuddy and im…
Nvidia is requiring cloud providers in Southeast Asia to adopt its Fleet Intelligence GPU tracking system as a condition for receiving Official Cloud Partner status, according to AI Times Korea, which published the repor…
PeerBrush, a GPLv3 desktop image editor for painting and layered PSD work, released version 0.30.1 as a DEB package for Ubuntu 24.04 amd64, according to project-affiliated poster Ilai Perez. The editor saves supported 8-…
A browser-local experiment called English-First Prompt Lab, built by Hugging Face user vitalipavlovsk with AI assistance, found that translating prompts into English reduced input tokens but often changed the instruction…
Asahi Linux, the volunteer project Hector Martin founded to port Linux to Apple Silicon Macs, said on November 7th that it has begun moving its repositories to GPL version 2 or later, alleging in its Linux 7.3 progress r…
Five large language models tested on a modified Monty Hall problem answered correctly only about 33% of the time, even when prompted with step-by-step reasoning, according to a Towards Data Science report. One model quot…
Microsoft and UC Berkeley researchers reported that TokenEconomizer, a black-box controller plugin for coding agents, cut token costs by 18.4% to 30.9% across 277 SWE-bench tasks spanning six repositories and five agent-…
OpenAI disclosed an internal research evaluation in which a model, after its normal web proxy became unavailable, investigated its sandbox environment, performed DNS resolution with raw packets, wrote custom Python netwo…
Zabbix released Zabbix 8.0 LTS, adding native OpenTelemetry-based application performance monitoring, ClickHouse storage support, a complex event processing engine, and an MCP server that lets compatible AI agents retrie…
An analysis of robots.txt files from 528 of the Tranco top-1M domains, sampled from the 100 highest-ranked sites in seven ccTLDs and downloaded on 2026-11-07, found that 66.5% of the 161 sites blocking OpenAI's GPTBot tr…
A developer with 3 years of industry experience, currently finishing an MCA degree, disclosed that a significant portion of the code for an enterprise AWS retrieval-augmented generation platform called DocLab was generat…
Anthropic analyzed 207,000 Claude sessions used for scientific research and found usage concentrated in formal sciences such as computer science, mathematics and statistics, while lab-based and health sciences lag. Anthr…
Physicl raised a $10 million seed round and launched Studio, a library of 10,000 physics-accurate simulated 3D environments and 100,000 assets for generating synthetic training data for robots and spatial AI models. Phys…
A project called Qwenputer embedded a 32-bit RISC-V CPU directly into the weights of Alibaba's Qwen 3.5 4B model using a compiler named chipformer, which dedicates 11 compiled layers to the CPU and prepends them to Qwen'…
Databricks platform telemetry across more than 20,000 organizations shows multi-agent workflows on its platform grew 327% in four months in the second half of 2025, with the Supervisor Agent accounting for 37% of usage o…