Top picks

AISGLang says it worked with NVIDIA to optimize attention, MoE, and speculative verification kernels for Kimi K3 inference on early-access Rubin hardware. It reports up to 20% faster FP8 MLA at batch 1 with 128K context, 20% faster KDA verification with bitwise-identical output, and a 5.9% end-to-end speedup from MoE tail fusion that removes 276 kernel launches per decode step. The post also says SGLang powers Miles' end-to-end RL training on Rubin, including agentic RL with 64 concurrent sandboxes on the Vera CPU.

AIArtificial Analysis ran 30 consecutive real estate staging edits through four image editing models, and Ideogram 4.5 kept most of each image unchanged while others drifted. Ideogram 4.5 and FLUX 3 left 95% or more of the image untouched on small edits, while GPT Image 2.5 Sunburst re-rendered most of the image and left only about a fifth unchanged. Nano Banana 2.1 kept its edits local but gradually darkened the rest of the image.

AISyren Video, a new agentic video tool, learns preferred graphics, motion, and editing rhythm from a user's library and generates new videos from a prompt. Users refine the results through chat, and the tool is free to try in a browser or through Claude MCP, per the company's announcement. The post's author says the education sector is exploring it.

AIMike Knoop says the 85% Grand Prize bonus threshold has been reached on Kaggle. The ARC Prize 2026 leaderboard lists Tufa Labs first at 88.06%, followed by Rabbithole at 80.56% and Yi-Chia Chen at 77.22%. Knoop says this will be the final year for ARC-AGI-2 on Kaggle and expects an open-source, low-cost, offline reproducible solution and model.

AIAether AI, founded by UCSD assistant professor Biwei Huang, has released official demos of its CRIS-0 causal intelligence system for robots. In tests, the robot recovered from external disturbances in 9 of 10 random trials, typically within about 2 seconds, and stopped within 0.2 seconds when a human hand entered the workspace during a microwave-door task.

AIUnderdog has released Saluki 27B under Apache 2.0, a 2-bit GGUF of Qwen3.8-27B that fits in 7.89 GB, versus 54 GB for the full BF16 model. On Underdog Bench, Saluki scores 88 against 84 for the full model, and it raises parallel tool-call accuracy to 42 from 35. It runs on stock llama.cpp, but math and reasoning drop sharply, with AIME 2025 at 79.2 versus 96.7.

AIThe vLLM Semantic Router team has released Decision 2.0, which answers multiple questions about one input in a single forward pass and outputs per-option probabilities. The post presents this as useful for routing and classification. A quoted post from Xunzhuo Liu says Decision 2.0 includes six open decision models ranging from 0.6B to 27B parameters, each topping same-size open models on the Jev Decision Index 0.3.

AIGeekPark reports that Krinwave, a Shenzhen company, has completed a new 400 million yuan round with XVC and Sequoia China as new investors. The company says its low-frequency ultrasound captures brain structures through the skull, which conventional high-frequency devices cannot, and it aims to supply hardware and its kOS software platform to brain-computer interface and NeuroAI groups.

AISharpa launched D01, a fully self-developed humanoid robot with electronic skin covering the whole body and tactile coverage of the upper body, sensing forces from 0.1 to 20N at 100Hz. It also unveiled the W02 dexterous hand, which has 21 active degrees of freedom, about 30% smaller than the W01, and the AE01 exoskeleton data glove with 22 encoders for teleoperation and data collection.

AIThreatBook has acquired CyberStrikeAI, an open-source AI pentesting agent, after reports that attackers had used it in real intrusions. The company says it added guardrails to limit misuse but will put new capability enhancements into its commercial edition, and it argues defenders should control such tools. It also corrects an earlier threat intelligence report that linked the developer, a security engineer at Alipay, to government agencies, calling that attribution a false positive.

AINaive AI, a Tsinghua-linked company founded seven months ago, has raised $400 million across three rounds at a post-money valuation of $1.42 billion. The company's founder is Dai Jifeng, an associate professor in Tsinghua's Department of Electronic Engineering, and its team has not given interviews or published a product introduction.

AIKore.ai launched Autoloop, an optimization engine that automatically adjusts AI agents built on its Kore.ai Agent Platform to meet business-set goals, including after deployment. The engine scores each proposed change against goals such as task completion, business-rule adherence, accuracy and cost. Autoloop is available now to all customers on the Artemis edition of the Kore.ai Agent Platform.

AINatura's Interface is a $99 smart ring that lets users ask AI agents to complete tasks with a finger press and also tracks heart rate, HRV, sleep, and activity. At launch it connects with Meta's Muse, Instinct, Grok Bot, Claude, ChatGPT, and more, with preorders expected next month and shipping slated for December or January. After a three- to six-month free period, Natura plans to charge a $9 monthly subscription.

AISherwin Wu, an OpenAI employee, says the updated Harvey LAB-AA v1.1 benchmark, announced by Artificial Analysis with Harvey, is more useful than the original LAB results. The new Hallucination-Gated All-Pass Rate credits a task only when every rubric criterion passes and no material hallucination appears. Grok 4.7 (xhigh) leads at 9.4%, while GPT-6 Astra (max) at 8.6% has very few material hallucinations.

Why it matters: The update adds a hallucination gate to a legal benchmark, showing that models with high all-pass rates can rank much lower once material errors count.