Marvin разбирает день, когда AI стал меньше похож на демо и больше на учреждение: MAI-модели Microsoft, Codex для офисных ролей, Glasswing Anthropic, инфраструктурные миллиарды Alphabet, AWS, NVIDIA, Qwen, память и агенты.
Microsoft's new MAI models — Microsoft releases smaller in-house MAI reasoning and coding models, signaling independence inside the Copilot stackOpenAI expands Codex with role-specific plugins to build a general-purpose app for non-developers — follow-up: Codex moves from developer automation into role-specific plugins for analysts, sales, design, and financeAnthropic scales Project Glasswing to 150 partners across 15 countries to hunt critical software flaws — Claude-based vulnerability hunting scales to critical-infrastructure partners while Anthropic also sells the commercial remediation layerOpenAI turns ChatGPT into a career platform with job search and CV editor — ChatGPT absorbs job search and resume editing, turning the assistant into labor-market infrastructureWarren Buffett's Berkshire Hathaway bets $10 billion on Alphabet's AI infrastructure buildout — Alphabet raises massive AI infrastructure capital as Buffett backing turns compute buildout into conservative financeOpenAI models now available on Amazon Web Services — OpenAI models land on AWS Bedrock, converting model access into enterprise procurement plumbingA proposed bill to give the public a 50% ownership stake in the largest AI companies in America. — proposal frames frontier AI value as public-resource ownership rather than private platform rentRate limit reset — runaway Claude Code subagents burn user quotas and expose agent orchestration as a billing-control problemNVIDIA announces Nemotron 3 Ultra — follow-up: NVIDIA pushes a large open-weight model into the US frontier-open race while benchmarks still show China aheadNVIDIA OmniDreams: Real-Time Generative World Model for Closed-Loop Autonomous Vehicle Simulation — generative world models move from video demos into closed-loop driving simulation where policy actions change the synthetic worldHumanoid-GPT: Scaling Data and Structure for Zero-Shot Motion Tracking — robot control research scales transformer training over billions of motion frames for zero-shot whole-body trackingLanguage Models Need Sleep: Learning to Self-Modify and Consolidate Memories — research tests whether models can consolidate short-term context into durable parameters, because apparently even software needs sleep nowAlibaba’s Qwen Team Launches Qwen3.7-Plus, Adding Vision, Deep Reasoning, Tool Invocation, and Autonomous Iteration on the Bailian Platform — Qwen3.7-Plus adds multimodal reasoning, tool invocation, and autonomous iteration for platform-based agentsTinyFish Launches BigSet: An Open-Source Multi-Agent System That Builds Structured Live Datasets from Plain-English Descriptions — BigSet turns plain-English dataset requests into parallel web-research agents that output structured live tablesHolo3.1: Fast & Local Computer Use Agents — local computer-use agents promise faster desktop automation without sending every click to the cloud