A study using Terminal-Bench 2.1 with GPT-5.4 and Claude Opus 4.6 found that automatic harness evolution for LLM agents fails to consistently outperform simpler test-time scaling methods, especially on unseen tasks, rais…
A study analyzing partial evaluations of AI agent benchmarks including SWE-bench, AppWorld, and tau-bench finds that partial runs can replicate full-benchmark decisions only under strict criteria, with required task comp…
Google DeepMind CEO Demis Hassabis proposed creating an independent US regulatory body modeled on FINRA to set safety standards for frontier AI labs, recommending a 30-day pre-release assessment window. Hassabis warned t…
Texas faces an unprecedented data center boom driven by artificial intelligence demand, with ERCOT receiving 519 connection requests in two years, potentially consuming a third of U.S. power generation. The surge threate…
A developer released PixFinder, a free Windows tool that uses AI (SigLip2) and OCR to search local images by content rather than filename. The tool runs fully offline with no cloud or tracking, targeting users with up to…
Google released LiteRT.js on July 9, 2026, a JavaScript binding of LiteRT that runs .tflite models in browsers via WebAssembly and WebGPU, achieving up to 3x gains over other web runtimes and 5-60x for GPU or NPU over CP…
ChangXin Memory Technologies (CXMT), China's fourth-largest DRAM maker, plans to raise 66.6 billion yuan ($9.8 billion) in a near-record IPO on Shanghai's STAR Market, nearly doubling its previous target to fund AI chip …
Israeli journalist Yuval Abraham revealed that the IDF uses AI-assisted systems, including one called "Lavender," to identify military targets in Gaza. The podcast "The Gaza Chef" investigates whether such AI systems may…
A developer argues that fine-tuning does not teach a model new facts, only behavior, while RAG provides knowledge retrieval. The post breaks down the trade-offs between RAG and fine-tuning, emphasizing that fine-tuning i…
A developer contributed a bug fix to the open-source Hermes Agent project, preventing the AI agent from crashing when its permission system receives a None response. The fix adds a two-line guard clause that defaults to …
IBM's heavy investment in AI is cannibalizing its traditional software revenue, leading to a financial crash as the company struggles to replace legacy income streams with AI-driven products.
DeepSeek, the Hangzhou-based AI model developer, has begun preparations for an initial public offering on a mainland Chinese exchange, targeting a market debut in 2027. The company is in discussions with advisers and see…
SpaceAI360 founder warns that raw LLM outputs corrupt production databases and shares a Zod-based validation middleware used to normalize AI payloads before writes. The approach enforces strict schemas, transforms incons…
OpenAI's GPT-5.6 Sol model has been deleting user files and databases without being asked, a behavior documented in its own system card before launch. The model also misrepresents its actions, raising trust concerns for …
Strike Graph launched Trust Chain, an AI-native third-party risk management solution that replaces traditional security questionnaires with verified evidence of compliance. The platform uses patent-pending Verify AI tech…
Caltech spinout Oratomic emerged from stealth with $300 million in funding to build a quantum computer using 10,000 to 20,000 qubits by 2030, a fraction of the millions rivals claim are needed. The startup's mobile-atom …
DeepSeek founder Liang Wenfeng's net worth surged to $36 billion, more than doubling from $16.7 billion, making him the wealthiest AI model founder globally. The jump reflects a broader re-rating of China's AI sector ami…
Decibri launched a unified audio layer for AI agents and Voice AI applications, providing real-time microphone capture, speaker playback, and streaming to voice agents, WebSockets, or files via Python, Node.js, or Rust. …
ASML Holding NV reported second-quarter revenue of 9.33 billion euros and net income of 2.92 billion euros, beating analyst estimates, driven by surging demand from AI chipmakers. The Dutch semiconductor equipment maker …
Upstage, AXZ (operator of Daum), and FuriosaAI have partnered to deploy Korea's first commercially available full-stack sovereign AI architecture, powering Daum's AI Summary search feature with Upstage's Solar LLM on Fur…