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.
On the product front, Google research lead Logan Kilpatrick announced Veo 3 and the Gemini Embedding model are now accessible via API, alongside a free Gemini Pro subscription for Indian students and a preview of Gemini 2.5 Pro with Deep Search for Pro and Ultra subscribers. Phil Schmid said Gemini CLI has merged over 90 contributions, released a public roadmap, and is adding IDE and agent features. Comet lead Arav Srinivas showcased on-the-fly email and calendar cards that let users customize drafts and join meetings from generated UI.
On the AI tools front, HubSpot co-founder Dharmesh Shah praised Perplexity’s Comet AI browser as a mini customizable computer, running client-server compute and supporting local models. Additionally, Claire Vo demonstrated vibecoding, spinning up a Next.js/TypeScript/Sanity CMS scaffold in minutes for non-frontend coders.
On the PM side, Aakash Gupta noted that Google’s new interview format is being copied industry-wide and outlined three variants candidates should prepare for. Product discovery coach Teresa Torres offered a guide comparing open and closed ecosystems, helping PMs choose the model best aligned with their business goals.
In industry news, OpenAI CEO Sam Altman said the company will exceed one million GPUs online by year-end and challenged his team to pursue 100× growth. Product advisor Lenny Rachitsky highlighted an interview with Anthropic co-founder Ben Mann, who left OpenAI over safety concerns, forecast a 50% chance of superintelligence by 2028 with up to a 10% catastrophic-risk chance, defined an Economic Turing Test for transformative AI, and detailed Anthropic’s constitutional AI and self-critique via reinforcement learning.
Turning to engineering workflows, LaunchDarkly tech leader Zach Davis described centralizing AI agent rules and docs in one monorepo directory covering style guides, TypeScript essentials and accessibility to ensure consistency across agents like Cursor and Devon. They route frontend test logs into files, use Claude to cluster over 1,200 warnings and generate tiered AI task checklists, systematically reducing tech debt and noisy logs. They also trained a custom GPT on interview rubrics and past scorecards to generate Slack-ready feedback, maintaining a consistent hiring bar.
In a coding walkthrough, engineers set up a Grock MCP server to fetch trending AI posts via live search, then ran a cloud code script to capture date-stamped summaries and screenshots. They assembled a 60-second AI news video by generating a script, adding AI voiceover and images, stitching clips with ffmpeg, adding background music and automating uploads—then applied the pipeline to top Reddit posts.
That's a wrap on today's GenAI PM Daily. Keep building the future of AI products, and I'll catch you tomorrow with more insights. Until then, stay curious!