Full-text search across 25531 articles. Combine with topic and date filters; results sorted by relevance.
A new essay from The Number Laboratories argues that recursive self-improvement in AI should be measured by how much time a model takes off the schedule for its successor, proposing an organizational definition where AI-…
Four frontier AI labs have dismantled or weakened their internal ethics and safety structures over the past two years, according to AI Weekly's Issue #523. OpenAI alone dissolved three safety teams—Superalignment (2024),…
A controlled study of 260 agent configurations by Google Research, Google DeepMind, MIT, and collaborators found that multi-agent systems improved performance by 81% on decomposable financial reasoning but degraded perfo…
A new analysis argues that AI spending only makes sense if customers buy roughly $1.2 trillion of tokens a year, rising to $2.5 trillion by 2031, and concludes the buyer is payroll, with no return on investment. The piec…
Google Gemini fell four places to rank 16 in Similarweb's July 2026 ranking of the world's most-visited websites, while ChatGPT held firm at rank 5. The drop follows Gemini's first traffic-share decline in over a year, a…
A court-records watch using CourtListener's RECAP mirror of PACER has identified 23 federal lawsuits naming a major AI company as a party, filed between 2026-07-13 and 2026-08-06, spanning product-liability, civil-rights…
Mark Zuckerberg's Meta published a 6,500-word manifesto titled "The Future is for Everyone" arguing that superintelligent AI will primarily drive invention rather than automation, citing potential breakthroughs in drug d…
GitHub Copilot, Google Antigravity, JetBrains Air, Kiro, Zed, and Zenflow are among the AI-powered development environments reshaping software creation, shifting developers' focus from writing code to reviewing and testi…
DeepSeek released V4-Flash-0731 on July 31, 2025, and the smaller model now outperforms the larger V4-Pro-Preview across all nine published agentic benchmarks, including Terminal Bench 2.1 (82.7 vs. 72.1) and DeepSWE (54…
OpenAI has confirmed the existence of its next major model, Astra, which has solved 10 major open math problems and developed advanced cyber capabilities, prompting the company to pause some internal development and impl…
RagLeap has released RagLeap Core, an MIT-licensed, open-source RAG engine that powers its self-hosted AI business manager, allowing users to run a document-grounded chat system on their own infrastructure with their own…
A fundamental flaw in how large language models identify instructions makes them impossible to fully secure against attacks, researchers argue in a paper presented at the International Conference on Machine Learning. By …
A coalition of technology companies published a letter defending open-weight AI models, urging the US government to avoid premature restrictions on downloadable models, four days after an Axios report that the Trump admi…
Epoch and METR released MirrorCode, a benchmark for long-horizon programming tasks, finding that AI models like Opus 4.7 solved a task in 14 hours for $251 that would take a human 2-17 weeks. Across 25 target programs, 1…
A growing number of American techies and companies, including Mozilla CTO Raffi Krikorian and Coinbase, are adopting Chinese AI models like Moonshot's Kimi K3 and Z.ai's GLM-5.2 because they are cheaper and increasingly …
Sandbox-CLI, an open-source tool that runs coding agents like Claude Code and Codex inside disposable Docker containers with strict isolation, is now in public beta. The tool mounts only the user's project directory, blo…
The Trump administration is exploring restrictions on Chinese open-weight AI models after Moonshot AI's Kimi K3 release, but the policy risk is hitting US startups hardest because the models are already downloadable, cop…
Moonshot AI's Kimi K3, a 2.8-trillion-parameter open-weight model with a one-million-token context window and Mixture-of-Experts design, scores 57.11 on Artificial Analysis's Intelligence Index as of July 21, trailing Cl…
Moonshot AI's Kimi K3, a 2.8-trillion-parameter open weights model, is the largest ever built and its benchmark performance rivals US frontier models, reigniting debate over China's AI competitiveness. The Register repor…
SpaceX short interest has surged from 6% to 32% of the public float in one month, representing $25 billion in bearish bets, as the company sets an August 4th earnings date that could unlock up to 911.5 million insider sh…