ChatGPT Rolls Out Record Mode
Today's curated insights on AI product management from X/Twitter across 60+ expert sources and YouTube channels.
ChatGPT Rolls Out Record Mode
From X
AI Product Launches & Updates
Record mode rollout in ChatGPT: OpenAI Team @OpenAI announced Record mode is rolling out to Pro, Enterprise, and Edu users, available on macOS desktop app.
Claude Code connects to remote MCP servers: Anthropic Team @AnthropicAI announced that Claude Code can now pull context from your tools directly into Claude Code with no local setup required, available to all users (What’s new).
Video FPS support in Gemini API: Logan Kilpatrick @OfficialLoganK shared that they shipped video FPS support in the Gemini API, enabling dynamic frame rate adjustments for enhanced video use cases.
AI Tools & Applications
Agent Evals in LangGraph Studio: LangChain Team @LangChainAI announced no-code evaluator support, letting you run agents over LangSmith datasets and apply evaluators in LangGraph Studio without writing a line of code.
AG-UI integration for agent frontends: Julien Chaumond @llama_index launched official AG-UI support, making it simple to build agent-powered frontends with @copilotkit and zero boilerplate.
Product Management Insights & Strategies
“AI drive” for self-organizing assets: Arav Srinivas @AravSrinivas proposed an “AI drive” inside Perplexity to store code, tables, charts, markdown, etc., in a self-organizing, searchable space.
Maximizing AI impact for PMs and founders: Shreyas Doshi @shreyas noted that there are far more impactful ways for product managers and founders to leverage AI—beyond current practices—by rethinking context and workflow design.
Anyone can become a PM in 2025: Aakash Gupta @aakashg0 highlighted that PM job openings are up 54% since early 2023 and shared a guide for breaking into PM roles from any background.
AI Industry Developments & News
OpenAI Podcast debut: OpenAI Team @OpenAI launched the OpenAI Podcast series, kicking off with @sama and @andrewmayne discussing AGI, (wen) GPT-5, privacy, and future directions.
Meta’s $100M talent poaching: Rowan Cheung @rowancheung reported that Meta is offering $100M signing bonuses to recruit OpenAI engineers—though none of OpenAI’s top talent have accepted.
The next great distribution shift: Brian Balfour @bbalfour shared an essay on a $100B distribution shift poised to reshape platform economics in the AI era.
From YouTube
How to make internet money
Greg Isenberg • June 18, 2025
Greg Isenberg interviews Jonathan Courtney about his $500K/month self-liquidating funnel—using a $6.99 paid book, checkout bumps, and one-time offers to generate $280K upfront and $300K backend in two weeks and book calls for $14,300 trainings—and then outlines scaling ideabrowser.com via a live webinar funnel.
Key Takeaways:
- Jonathan Courtney’s funnel, built in ClickFunnels, converts 58% of cold traffic on a $6.99 lead-magnet book and pulled in $235K in upfront sales with a $68.85 average cart value within its first two weeks.
- By layering checkout bumps (a $29 video course and up to $80 asset bundles) and conditional one-time offers, the funnel self-qualifies buyers and has generated 280+ call bookings for a $14,300 in-person training since its May 26 launch.
- Greg recommends ideabrowser.com run a 10-day live webinar titled “Behind the scenes of how I built a zero to $1M/mo diversified holding company,” promoted via email and social posts to drive sign-ups for its $299/year and $1,000/year plans.
New course with Meta! Enroll in "Building with Llama 4" 🦙
Deeplearning.ai • June 18, 2025
This video introduces Building with Llama 4, a Deeplearning.ai course taught by Amit Sangani in partnership with Meta, covering Llama 4’s new mixture-of-experts models, multimodal and long-context capabilities, and hands-on lessons using the Llama API for image reasoning, code analysis, prompt optimization, and synthetic data generation.
Key Takeaways:
- Llama 4’s MOE architecture includes Maverick (400 B parameters, 120 experts, 17 B active) and Sculpt (109 B parameters, 16 experts, 17 B active) to improve serving efficiency.
- The models add multimodal features—image reasoning and grounding—and support extended context windows up to 1 M tokens for Maverick and 10 M tokens for Sculpt.
- The six-lesson course by Amit Sangani covers Llama API usage, image bounding-box tasks, large GitHub repo analysis, prompt optimization, and synthetic data generation.