OpenAI launched GPT-5.5 (codenamed 'Spud'), reclaiming the frontier lead with new reasoning and coding capabilities. Meanwhile, the White House accused Chinese labs of industrial-scale model distillation, and Anthropic's 80K-worker survey revealed that AI's biggest productivity beneficiaries are also the most anxious.

This issue of The Rundown: Anthropic's locked-down cybersecurity model Mythos leaked to a Discord group days after release; SpaceX locked in a $60B acquisition option on AI coding startup Cursor; OpenAI launched Workspace Agents for teams; plus a two-step dictation strategy for better docs.

OpenAI launches ChatGPT Images 2.0 with 'thinking' capabilities, topping text-to-image leaderboards. Meta is reportedly logging employee keystrokes and mouse activity to train AI agents. Google releases Deep Research Max agents. A new guide shows how to build a daily command center with Claude Live Artifacts.

Google co-founder Sergey Brin personally forms a strike team to chase Anthropic's coding edge. Moonshot AI open-sources K2.6, an agentic coding model rivaling GPT-5.4 and Gemini 3.1 Pro. Adobe launches CX Enterprise agentic platform for businesses.

Anthropic launches Claude Design with Opus 4.7 vision model; OpenAI loses three senior execs in one day.

Canva CPO Cameron Adams joins The Rundown for an exclusive conversation on how Canva AI 2.0 bridges the gap between words and vision.

OpenAI transforms Codex into a 'superapp' with background computer use and memory, while Anthropic releases Claude Opus 4.7 to top GPT-5.4 in agentic coding, intensifying the rivalry between the AI giants.

This week: Allbirds reinvents itself as a GPU rental business, stock surges 600%; Google launches native Gemini Mac desktop app; Notion's built-in Claude agents can audit business workflows; Snap cuts 1,000 jobs as AI now writes 65% of new code.

OpenAI launches GPT-5.4-Cyber with wide defender access, Nvidia releases open-source quantum AI Ising, and Anthropic redesigns Claude Code desktop.

Andon Labs puts an AI agent named Luna in charge of a real retail space with $100K budget and a credit card, letting it hire workers and manage a boutique. Stanford HAI reports AI adoption has surpassed 50%, but public trust hits record lows.

A 20-year-old was arrested for attacking Sam Altman's home with a Molotov cocktail. Altman responded by acknowledging 'justified' AI anxiety. Other updates include AI-driven drug side effect discovery and infographic generation tools.

Perplexity integrates Plaid to turn Computer into a personal finance hub, pushing ARR past $450M with a 50% monthly jump; Amazon discloses AI revenue for the first time; Oxford AI predicts heart failure five years early.

Meta's Superintelligence Labs releases multimodal reasoning model Muse Spark; HeyGen Avatar V solves AI avatar identity drift; Anthropic launches Claude Managed Agents; ElevenLabs Flows enables automated ad generation.

Anthropic 发布 Project Glasswing 网络安全联盟，展示未发布的 Mythos AI 模型。该模型在漏洞检测和编程能力上超越现有所有模型，但 Anthropic 认为其过于强大，仅向 12 家合作伙伴开放，不公开发布。...

Anthropic unveils Mythos AI, deemed too powerful for public release; China's Z AI GLM-5.1 tops coding benchmarks with open-source model; Anthropic locks in 3.5GW compute deal with Google and Broadcom, reaching $30B run-rate revenue.

Sam Altman 发表长文阐述 AI 发展与社会契约的关系，呼吁建立新的全球治理框架，强调 AI 安全的重要性。...

OpenAI proposes taxing AI profits, creating sovereign wealth fund, 4-day workweek; New Yorker publishes investigation into Altman firing; Meta's first superintelligence models ready to ship.

Anthropic requires OpenClaw users to pay separately; Netflix releases VOID physics-aware video editing; Granola brings AI notetaking to iPhone calls.

AI 工具正在改变创业游戏规则，让个人创始人能够创建估值数十亿美元的公司。AI 自动化降低了对大型团队的需求。...