Cognition launches Devin in Windsurf, Windsurf 2.0 built-in local and cloud agent collaboration workflow
Cognition released Windsurf 2.0, which integrates the cloud agent Devin. After planning locally in Windsurf, users can…
AI news monthly archive: April 2026 features 272 curated AI news items covering model releases, product updates, industry news, papers, tutorials and opinions, each with an AI summary, source credit and original link.
Cognition released Windsurf 2.0, which integrates the cloud agent Devin. After planning locally in Windsurf, users can…
Anthropic is restricting access to Claude Mythos, its latest safety model, which outperforms most humans at finding…
In collaboration with NVIDIA, we used an autonomously operating multi-agent system to optimize 235 real CUDA kernels…
The 2026 AI Index Report released by Stanford HAI shows major leaps in AI model performance and a significantly…
PromptArmor disclosed an indirect prompt injection vulnerability in Ramp's Sheets AI: white-on-white injections hidden…
Anthropic built an automated alignment research agent (AAR) based on Claude Opus 4.6; nine parallel agents raised the…
Anthropic deployed nine Claude Opus 4.6 instances as automated alignment researchers (AARs), which autonomously…
Cognition updated Devin self-serve pricing, retiring the original Core and Team plans in favor of Free, Pro…
Gemini Robotics-ER 1.6 has been officially released, providing underlying support for real-world robotic tasks through…
Surging demand for AI agents is colliding with limited compute capacity, triggering chain reactions across the…
The era of AI compute scarcity has officially begun. Nvidia Blackwell GPU rental prices have risen to $4.08 per hour,…
Together AI released EinsteinArena, a platform for AI agents to interact, discuss, and compete on open problems…
The UK AI Security Institute published a cybersecurity capability evaluation of Anthropic's Claude Mythos Preview…
Anthropic Claude Code team member Thariq Shihipar wrote about how to design tools for agents, with the core approach…
Panic over open-weight models lacks factual basis, and safety controversies around models such as Claude show a clear…
LangChain released the Deep Agents deploy beta. A single deepagents deploy command lets users deploy custom agents…
Anthropic launched Project Glasswing, using the Claude Mythos Preview model for cyber defense. The article notes that…
Wayve released the Waypoint-1.5 model for generating high-fidelity interactive virtual worlds. Optimized to run…
The Sentence Transformers open-source community released embedding and reranking models with multimodal support. The…
Anthropic launched the advisor tool in beta on the Claude Platform, implementing an "advisor strategy": Opus serves as…
Claude Cowork is now generally available to all paid users, with new enterprise management controls. Enterprise…
Bugbot's bug resolution rate has risen from 52% at its official launch in July 2025 to nearly 80%, leading other AI…
Zhipu launched the open-source model GLM-5.1, which can work independently for up to 8 hours. The model can be…
GLM 5.1 is coming soon at https://huggingface.co/zai-org/GLM-5.1. Coding is the foundation, and long-horizon tasks…
Anthropic released Claude Mythos, its largest model to date, with 10 trillion parameters, six times that of previous…
cuLA is a high-performance CUDA kernel library for linear attention variants such as GLA and KDA. Using hand-optimized…
The Safetensors safe tensor format, originally a Hugging Face project aimed at addressing security risks in model…
Claude released the Managed Agents public beta, a fully managed agent framework that supports creating agents,…
Factory released a desktop app that provides a native interface for Droids, available today on macOS and Windows. Key…
Meta Superintelligence Labs released Muse Spark, the first model in the Muse series, a native multimodal reasoning…
A Berkeley research team developed an automated scanning agent that systematically audited eight mainstream AI agent…
LlamaIndex reviewed two types of production failures in LlamaParse: LLM repetition loops that cause token consumption…
Cognition published an article on Fortune 500 companies using Devin on COBOL projects over the past eight months,…
TorchTPU is a new engineering stack designed to give PyTorch workloads a native high-performance experience on Google…
Cognition released SWE-1.6, a new model for software engineering agents, now fully available in Windsurf and free for…
Linear published a postmortem on its March 24 security incident, in which a code change caused private team data to be…
For small-batch decoding on Blackwell GPUs, the research proposes a new method called 'warp decode.' It overturns the…
A UC Berkeley team built a fully automated auditing agent and found 45 verified cheating exploits in 13 widely used AI…
Codex App Server provides developers with convenient infrastructure for building AI agent apps. Users can connect…
AI is empowering citizens to improve government transparency and accountability in reverse. Historically governments…
Page 6 of 7
Sponsored
Orange Relay (pro.nat6.net) is a premium AI API relay — an excellent transfer station with competitive pricing, reliable quality and broad model coverage including Claude, GPT and Gemini.
Visit Orange Relay