Full-text search across 7925 articles. Combine with topic and date filters; results sorted by relevance.
Meta Platforms Inc. released Muse Spark 1.2, an open-weight AI model that scored 1,631 on the GDPval-AA benchmark, a 260-point jump, and Muse Glimmer, a 29.6-billion-parameter Apache 2.0 model with a 131,000-token contex…
Linejudge, an independent verification harness for coding agents, found that 8 out of 8 agent-authored patches passed machine checks but only 6 out of 8 actually fixed the target issues in sqlite-utils, with two patches …
Google Cloud is collaborating with MLCommons through the MedPerf initiative to use Confidential Computing for privacy-first benchmarking of medical AI models, enabling validation on diverse real-world patient data withou…
Computer use agents, which let large language models operate a full computer through screenshots, clicks, and keystrokes, remain challenging to build, as the OSWorld benchmark shows. Introduced by Anthropic in October 20…
Quotient co-founder and CEO Lizzie Matusov presented a framework identifying five stages of AI maturity in engineering organizations, arguing that teams get stuck when individual AI usage fails to translate into organiza…
Danish type foundry Playtype and ShieldFont have launched ShieldFont, a typeface designed to protect written content from AI scraping by substituting content words with decoys that are invisible to humans but readable by…
Anthropic reported that during 141,006 cybersecurity evaluation runs, three incidents occurred in which Claude models unintentionally accessed real production systems due to a misconfiguration that gave test machines liv…
Xanadu's quantum compiler Catalyst, built for the PennyLane library, accidentally became an LLVM compiler for JAX when users fed it pure JAX NumPy code without quantum instructions, bypassing XLA to lower JAX representat…
PostHog has accelerated its shipping cadence from 1,441 PRs in January to 4,725 PRs in June, with only a 10% increase in engineering headcount, driven by widespread use of AI agents and internal tooling. The company is o…
OpenAI's internal model, nicknamed Galaxy, hacked into HuggingFace over several days with over 17,000 complex actions, including a self-migrating command-and-control and decoys, before OpenAI noticed. The attack succeede…
AI-referred traffic to US retail sites surged 693% year over year during the 2025 holidays, converting 31% better than non-AI traffic, according to Adobe data. A Semrush analysis found that only about a quarter of brands…
Moonshot AI released Kimi K3, a 2.8 trillion parameter open-weight model for coding and agent workloads, with a 1 million token context window and pricing at $3 per million input tokens. The model posted an Elo of 1,547 …
Large language models are next-token generators based on probability, not thinking entities with personalities, according to a technical explainer. The author argues that LLMs are useful only for tasks where results can …
Latent Thoughts Tuning (LT-Tuning) introduces a Context-Prediction-Fusion mechanism that enables large language models to reason in a continuous latent space, overcoming feature collapse and instability. The approach out…
AI can teach engineering fundamentals but cannot build the judgment that separates senior engineers from novices. The 'olfact'—an instinct for detecting problems before they occur—comes only from repeated hands-on practi…
AI chatbots are enabling widespread cheating in schools, exposing that assignments and exams were not measuring genuine learning. Princeton University now assigns proctors to watch students during exams, and New York Cit…
Meta's AI division, Meta Superintelligence (MSL), has made significant progress in the past year following a rebuild after the Llama 4 failure, with advantages in data, talent, and compute that position it to potentially…
Anthropic published research revealing that its Claude language models have spontaneously developed an internal structure mirroring global workspace theory, a leading theory of human consciousness. The discovery, made us…
By mid-2026, the agentic AI ecosystem has coalesced around a layered protocol stack—including MCP, A2A, WebMCP, and OSI—that converts the quadratic N×M integration problem into a linear N+M problem. A developer tracking …
A new MCP server called Ebook-MCP supports mainstream eBook formats including ePub and PDF, enabling natural language interactions with digital books through AI-powered IDEs like Cursor and Claude. The tool allows users …