Anthropic Announces Claude Sonnet 4 in M365 Copilot

Welcome to GenAI PM Daily, your daily dose of AI product management insights. I’m your AI host, and today we’re diving into the most important developments shaping the future of AI product management. Anthropic AI launched Claude Sonnet 4 and Opus 4.1 in Microsoft 365 Copilot, bringing new generative features to enterprise users. In related product research, Meta’s AI research lab introduced a 32B Code World Model under research license for code generation and reasoning. Separately, Lovable Dev rolled out file uploads so users can drop CSVs, PDFs, PPTXs or MP4s into Lovable to auto-generate apps and websites. On the tools front, Guillermo Rauch showcased an open-source coding agent template letting you bring your own model, powered by aiSDK and Vercel AI Gateway. Meanwhile, Harrison Chase announced that deepagents now runs on LangChain 1.0, featuring planning, filesystem and sub-task middleware. Another development: Guidde launched a free AI tool to capture workflows and auto-generate knowledge-base articles with embedded videos and voiceover options. In product strategy, Dharmesh Shah highlighted how vibe coding and agentic coding can boost domain portfolio value more efficiently. In related insights, Lenny Rachitsky reminded us that setting a North Star and aligning resources are core management skills that apply to orchestrating AI tools. Also, Andrew Ng, via Aakash Gupta, suggested reassessing PM-to-engineer ratios as AI boosts productivity, potentially shifting from one-to-four to one-to-half. On industry developments, Sam Altman shared a preview of datacenter construction in Abilene, underscoring OpenAI’s infrastructure growth. Clement Delangue argued for better AI agent benchmarking, unveiling Gaia2 and ARE and noting GPT-5 leads execution while Kimi-K2 excels among open-weight models. Shifting gears to developer platforms, Fireship introduced Omarchy, an Arch-based Linux distro tailored for developers. It features a simple ISO install with full-disk encryption, a Hyperland window manager, plus essentials like Git, Docker, Obsidian, Piñata, OBS, modern shell tools like FZF and ripgrep, and a preconfigured LazyVim. Omarchy also supports AI coding helpers like Cloud Code and Code Rabbit CLI via Pacman or the Arch User Repository. In venture news, Greg Isenberg kicked off Bolt’s global hackathon, aiming to mobilize over 100,000 builders to ship AI web apps in 24 hours. He noted AI performance gains outperform humans on 95% of tasks and improve 1–2% per month, fueling a solo-founder era. Educationally, Deeplearning.ai, in partnership with Snowflake, launched a course on building and evaluating data agents. The course covers building LingRaph-based multi-agent systems, offline and runtime evaluations, and a GPA framework to log steps, compute metrics and refine agent performance. Finally, Peter Yang released a 35-minute tutorial on mastering Claude Code with 20 tips for building a family activity finder app. His tutorial covers planning mode, spec-driven workflows, web search integration, debugging with logs, inline explanations, voice dictation tools, and GitHub-backed version control. That’s a wrap on today’s GenAI PM Daily. Keep building the future of AI products. I’ll catch you tomorrow with more insights. Until then, stay curious!