Google Cloud's new Discovery Bench framework exposes alarming flaws in AI benchmarking, revealing that a single vague word can crash an agent's performance score from perfect to zero, while also uncovering broken ground truth and mislabeled data in widely trusted benchmarks — raising urgent questions about whether the AI industry …
OpenAI's new GPT-5.6 model is now powering Microsoft 365 Copilot, bringing cutting-edge AI capabilities to Word, Excel, and PowerPoint, enabling users to draft documents, analyze data, and build presentations faster and more efficiently than ever before.
OpenRouter launches the Pareto Router, a free tool that automatically picks the most cost-efficient coding AI model based on a user-defined quality threshold, with session stickiness to ensure consistent results across multi-turn conversations.
Prisma ORM is taking the developer world by storm, amassing 774,000 users and 47,000 GitHub stars with its type-safe database toolkit that supports seven major databases including PostgreSQL, MongoDB, and MySQL for Node.js and TypeScript developers.
OpenAI acquires Northslope, a Palantir-backed applied-AI firm, deploying hundreds of engineers directly inside client businesses to dominate enterprise AI implementation as competition shifts from model performance to real-world adoption.
OpenAI acquires applied AI firm Northslope to expand its team of engineers who embed directly with enterprise clients, signaling a major industry shift where AI companies are now competing head-to-head with traditional consulting firms to help businesses adopt and scale AI.
A new study reveals that using JSON mode for AI agents causes smaller models to fail up to 60% of the time and costs up to 11x more than traditional CLI arguments due to shell escaping failures, with developers urged to avoid rewriting CLIs for agents until running measurable experiments.
A sample of 1.2 million Claude Cowork sessions reveals business operations tasks dominate usage at 33.4%, nearly four times the 8.7% share held by software development, with roughly half of all usage falling into cross-role, supportive tasks rather than core job functions.
DoorDash launches DashBench, a rigorous AI code review benchmark that replays historical pull requests to surface real developer findings, revealing that multi-stage architectures dramatically outperform single-pass models and that trustworthy AI evaluation demands triangulating multiple imperfect signals rather than relying on any one metric.
Uber burns through its entire 2026 AI coding budget in just four months, with COO Andrew Macdonald openly questioning whether the skyrocketing spending is actually delivering value, admitting a clear link between AI tool usage and meaningful consumer features has yet to be established.