A custom C/Metal inference engine called Flash-MoE is now running a massive 397-billion parameter AI model on a standard MacBook Pro with 48GB RAM, streaming 209GB directly from SSD at 4.4 tokens per second — with 58 documented experiments revealing that Apple Silicon's unified memory architecture defies conventional optimization wisdom.
Microsoft launches MAI-Image-2, a powerful next-generation text-to-image AI model debuting at third place on the Arena.ai leaderboard, offering hyper-detailed photorealism, legible text generation, and accurate skin tones, now available in preview on the MAI Playground with Copilot and Bing Image Creator rollout underway.
Cursor's newly released Composer 2 coding model outperforms Anthropic's Claude Opus 4.6 on Terminal-Bench 2.0 benchmarks while costing up to 90% less, thanks to a breakthrough 'self-summarization' training technique that compresses context windows and cuts compaction errors in half.
OpenAI is racing to deploy an autonomous AI research intern by September 2025, with ambitions to build a fully automated AI researcher by 2028, even as major safety and misuse concerns remain unresolved.
MIT researchers unveil a breakthrough uncertainty quantification method that detects overconfident AI responses by measuring disagreement across multiple similar LLMs, significantly reducing hallucinations and outperforming existing approaches across 10 tasks including math reasoning and question-answering.
GPT-5.4 launches with major frontend design upgrades, including stronger image understanding, native computer use via tools like Playwright, and a dedicated Codex Frontend Skill that produces premium, production-ready interfaces with best practices like full-bleed layouts and intentional motion built in by default.
Cursor launches Composer 2, a powerful new coding model that matches Anthropic's Claude Opus 4.6 on key benchmarks while costing 10 times less at just $0.50 per million input tokens, signaling a major shift in the AI coding tools market.
Baidu launches the Qianfan-VL series of vision-language models ranging from 3B to 70B parameters, alongside a new 4B Qianfan-OCR model featuring 'Layout-as-Thought' reasoning that tops major benchmarks, supports 192 languages, and is built for enterprise document understanding and visual reasoning tasks.
Anthropic has surged to capture 73% of enterprise AI spending among first-time buyers, up from a 50/50 split with OpenAI just 10 weeks ago, forcing OpenAI to reportedly consider abandoning its consumer-focused strategy to compete in the lucrative enterprise market.
AI coding tools like Claude Code and Cursor are rapidly transforming software engineering, making coding accessible to novices while tech leaders warn this automation is just the beginning, with Anthropic dominating the space and capturing over 73% of enterprise AI spending as white-collar knowledge work across all industries faces sweeping …