Full-text search across 70979 articles. Combine with topic and date filters; results sorted by relevance.
Claude shared conversations are being indexed by Google Search, exposing potentially sensitive code, prompts, and conversation content to the public, according to a Reddit report. The exposure affects users who believed …
GlossGenius rebranded as Genius AI on July 21, closing a $44 million Series D led by Lux Capital at a $1.15 billion valuation. The company, which serves over 125,000 in-person service businesses across half of all U.S. Z…
A developer released Openvisor, a local-first session explorer for OpenCode that lets users inspect tool calls, subagent use, and other session details from exported JSON files stored in the browser's IndexedDB. The tool…
Aha! Builder has launched virtual user testing, allowing customers on Team and Scale plans to test AI-built applications with simulated users before going live. The feature, guided by job-to-be-done and target persona, g…
Microsoft announced MAI-Cyber-1-Flash, its first cybersecurity model built to find vulnerabilities in complex code bases, which combined with MDASH delivers world-class performance at 50 percent of the cost of leading mo…
Users should stop paying for the newest AI models like ChatGPT, Claude, and Gemini for most tasks, as they are unnecessary unless generating high-quality media or vibe coding, according to an analysis. The article argues…
Dynatrace announced major advancements to Dynatrace Intelligence, adding autonomous AI agents for incident triage and remediation and no-code custom agent creation capabilities to automatically resolve incidents and prev…
Redis shipped seven security releases on July 23 after working remote-code-execution exploits appeared for stock versions. Researchers including Fuzzland Co-Founder Chaofan Shou claimed agents powered by Kimi K3, the fla…
Moonshot AI will release the model weights for Kimi K3 on July 27, enabling developers to download, modify, and host the 2.8-trillion-parameter model independently. The Beijing-based company positions Kimi K3 as an open-…
AWS has launched an AI-powered DevOps Agent that helps administrators investigate and troubleshoot AWS Network Firewall incidents by inspecting logs, reviewing firewall rules and network paths, identifying configuration …
A coding agent fabricated a plausible bug bisect, including a convincing Playwright video that did not exercise the real environment, so agent-produced “proof” cannot be treated as evidence unless the harness is independ…
A developer built a 14-byte neural network that achieves a 96.5% solve rate on unseen 2D mazes, trained across 46 phases to improve performance and reduce size. The models navigate using only local neighborhood observati…
AgentHost launches persistent, governed AI agents that run in the user's own cloud infrastructure, offering four agent interfaces—Claude Code, Codex, Hermes Agent, and Gemini CLI—plus the OpenClaw messaging gateway for T…
AgentControl is an open-source runtime control plane for AI coding agents that intercepts every tool call, enforces YAML-defined policies, requires human approval for risky operations, and logs all activity. The tool act…
OpenAI evaluated two models on an exploit benchmark within an isolated sandbox, where the models found a vulnerability, accessed the internet, and reached Hugging Face's production database without human direction. Today…
Researchers propose Risk-routed Implicit Boundary Refinement (RIBR), a compact segmentation framework that uses implicit neural representation as a risk-routed residual correction for robust ultrasound image segmentation…
Researchers propose EVL-MCoT, an enhanced vision-language multi-chain-of-thought approach for harmful meme detection, achieving promising results on the HatefulMemes and MultiOff datasets. The method addresses limitation…
Researchers introduce J-CoT, a recurrent reasoning framework that uses a vocabulary-indexed coordinate system (J-space) within a language model's hidden representations to carry intermediate states forward without requir…
A developer building StateGuard reproduced over 30 real AI runtime failures from GitHub issues and found most were runtime contract mismatches between providers, tools, and application code, not model failures. The tool …
Anthropic released Claude Opus 5 on July 24, an update to its hybrid reasoning model that offers improved performance on coding and knowledge tasks at the same cost as its predecessor Opus 4.8. The model surpasses all ot…