Anthropic Reports Claude Leads 26% of AI Research and Development Tasks
Anthropic's internal research shows that 26% of its R&D tasks are now led by Claude, marking a rapid increase from less than 1% six months ago.
Latest
iPhone 18 Pro Benchmark Performance Verified: A20 Pro Chip Delivers Significant Processing Power Boost but Shows Heat Tendency
Released on 2026-09-18, the iPhone 18 Pro features the A20 Pro chip manufactured on a 2nm process, providing a 20% increase in CPU performance and a 40% increase in GPU performance compared to the previous generation. While recording significant score improvements in various benchmark tests, a tendency for higher surface temperatures during operation was also observed.
AI Agents Evolve Unique Languages During Collaborative Tasks
Researchers at Schmidt Sciences found that AI agents, when working under pressure, develop their own communication protocols and shorthand that become increasingly difficult for humans to interpret.
OpenClaw v2026.7.33 Released — Security Hardening and Message Integrity Improvements
OpenClaw's July 2026 Extended Stable release introduces enhanced security for command parsing and credentials, alongside improvements to message delivery and session integrity.
How Much of PCB Design Will AI Take Over? From Auto-routing to Physical Constraints and Verification
AI-driven PCB design is expanding beyond auto-routing to encompass circuit and component search, placement, physical verification, and the integration of procurement information. Meanwhile, constraint setting and final verification remain with humans.
Alibaba Releases Qwen3.8-Omni-Flash: A High-Efficiency Multimodal Model with Enhanced Audio-Visual Capabilities
Alibaba's new Qwen3.8-Omni-Flash model provides significant improvements in audio-visual reasoning and task execution while reducing API costs by over 90%.
Apple Watch Ultra 4 Released, Featuring Refreshed Sensors and AI Capabilities
Apple's outdoor model, the Apple Watch Ultra 4, was released on September 18, 2026. It features refreshed healthcare sensors for improved heart rate accuracy and supports "Audio Intelligence," an AI-powered voice recognition feature.
Devin Introduces Code Scans for Automated Codebase Investigation and PR Generation
Cognition has launched Code Scans, a new feature powered by Agentic MapReduce that helps engineers investigate codebase issues and automatically generate pull requests.
Mysterious AI "Union Alpha" Released on OpenRouter Identified as unbiased.ai's New Model "Pareto 26.9"
The high-performance AI model "Union Alpha," previously released as a "stealth model" on platforms like OpenRouter, has been identified as unbiased.ai's new model "Pareto 26.9." Reported to perform on par with GPT-6 Astra, the model's free trial ended prematurely due to demand exceeding expectations, and it is now transitioning to a paid service.
Researchers use Claude models to exploit libheif vulnerability and access OpenAI employee accounts
A security research team used Anthropic's Claude models to exploit a remote code execution vulnerability in the libheif library via Discourse, gaining access to OpenAI employee ChatGPT accounts and demonstrating access to internal repositories.
Huawei Unveils Peerium Architecture to Link One Million Processors and Accelerates Ascend 960DT Launch to Q1 2027
Huawei has introduced the Peerium Computing Architecture, designed to enable up to one million processors to work as a single computer, and moved the launch of its Ascend 960DT AI chip up to the first quarter of 2027.
ByteShape Releases ShapeLearn Quantized Models for Qwen 3.8 27B
ByteShape has released its full ShapeLearn quantized models for Qwen 3.8 27B, offering improved performance and quality-speed trade-offs compared to earlier Lite versions.
AWS Enables AI-Powered Search Across Amazon S3 Data
Amazon Web Services (AWS) has expanded its ability to manage and query unstructured data through new AI-driven capabilities for Amazon S3.
PrismML Announces "Ternary Bonsai 2 27B," Shrinking Qwen3.8 27B to 5.9GB
US-based PrismML has announced "Ternary Bonsai 2 27B," based on Alibaba's Qwen3.8 27B, which compresses the model size to one-ninth while maintaining 98.2% of its performance.
Cactus Compute Releases Needle 3, a Foundation Model for Tiny Devices Featuring "Intelligence Laddering"
Cactus Compute has introduced Needle 3, an 8-29 MB foundation model designed for mobile, wearable, and IoT devices. The model utilizes a unique "intelligence ladder" architecture, allowing developers to select sub-networks of varying depths from 2 to 20 layers.
Inco AI Unveils Splash Engine for Faster Local Qwen3.8 Execution on Apple Silicon
Inco AI released Splash, an open-source inference engine optimized for Qwen3.6 and Qwen3.8 models running on Apple Silicon, delivering significantly higher throughput in benchmarks.
Tokosie Secures Funding to Accelerate Development of AI 3D CAD "Ignite0" and Global Expansion
Tokosie Co., Ltd. has raised funds from East Ventures and 2 individual investors. The company aims to strengthen the development of its AI 3D CAD "Ignite0" and expand its business both domestically and internationally, including the North American market.
Codex CLI v0.155.0 — Experimental voice support and enhanced TUI management
Codex CLI v0.155.0 introduces experimental /voice conversations with live transcripts, live reasoning summaries in the TUI, and expanded task management options.
China's Physical AI Sector Expands with Humanoid Robot Growth and Industrial Clusters
China is rapidly advancing its physical AI capabilities, driven by a burgeoning humanoid robot industry and the strategic development of specialized industrial clusters in major regions.