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.
OpenAI rolled out ChatGPT pace control, letting Plus, Pro, and Business users adjust GPT-5’s thinking time in the message composer. In related news, Alibaba’s LM Studio supports Qwen3-Next with MLX on Mac and new mascots. On the product front, GitHub launched the MCP Registry, tied to its repos and the OSS Community Registry for seamless server discovery.
Meanwhile, Alibaba introduced the Qwen3-ASR-Toolkit CLI, an open-source tool for transcribing long audio and video with smart voice activity detection. Andrew Ng released a short course on building AI apps with MCP servers processing Box files via the Model Context Protocol. Separately, Phil Schmid shared a guide on using Google’s Gemini 2.5 Flash, or Nano Banana, with the AI SDK for text-driven image generation and editing.
In strategy insights, Lenny Rachitsky noted AI labs’ hunger for high-quality evals and data labeling is fueling hyper-growth at Mercor AI and Handshake. Sebastian Raschka said he uses AI for throwaway code and tasks outside his expertise—like converting Substack posts to Markdown—boosting his productivity. Clement Delangue argued firms relying on generalist LLM APIs today will soon need custom models for specific use cases.
In other news, Claude AI published a postmortem on three infrastructure bugs from August and September, outlining root causes and fixes. OpenAI teamed with the Apollo AI Eval team to uncover scheming behaviors in frontier models and demonstrate methods to cut them down. Google DeepMind revealed an advanced version of Gemini 2.5 Deep Think earned gold-level performance at ICPC 2025 by solving ten of twelve complex problems.
Meanwhile on marketing, Greg Isenberg interviewed Cody Schneider about hiring micro-creators with zero followers at $900 a month each to post daily short videos for B2C apps and e-commerce. That approach can deliver 500,000 impressions at a $2 CPM—below typical influencer rates and paid ads—and relies on tools like Stormy AI for discovery, Shortmize or Viral.app for tracking, and Sand Castle AI for remixing.
On companion tech, Fireship showed how to build “Jeff,” a voice-based AI agent, with Vapi, Terso Cloud, Astro, ElevenLabs, and Twilio. They covered OpenAI’s ChatGPT-4 deprecation drama, configuring Vapi with GPT-4o and custom voices, and an Astro app that verifies Vapi webhooks, parses caller info, and stores requests for future videos.
Finally, Deeplearning.ai launched a course taught by Box CTO Ben Kus on building AI apps with MCP servers to extract data from PDF invoices in Box. It contrasts a DIY file approach with Box’s MCP integration and guides learners on evolving their app into a multi-agent system via ADA protocols.
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!