OpenAI Launches Branched Chats on Mobile
Today's curated insights on AI product management from 100+ sources across X, LinkedIn, and YouTube.
OpenAI Launches Branched Chats on Mobile
From X
AI Product Launches & Updates
Branched chats on mobile: OpenAI @OpenAI announced branched chats are now available on iOS and Android, with multi-threaded conversation support.
Multi-currency support: Sam Altman @sama highlighted multi-currency support, expanding global payment options.
Qwen Code v0.5.0 release: Alibaba Qwen @Alibaba_Qwen announced VSCode integration, native TypeScript SDK, and smart session management.
AI Tools & Applications
Secure AI coding agents: Llama Index @llama_index shared AgentFS and LlamaParse to control filesystem access and document understanding in AI coding agents.
Agent debugging with LangSmith: LangChain AI @LangChainAI demonstrated using LangSmith to define deep agents in a single file, design system prompts, and build an email triage agent.
NotebookLM for PRD critique: Nuri Janian @nurijanian shared how NotebookLM’s audio overview can be used to critique PRDs for faster feedback.
Product Management Insights & Strategies
Value of coding agents: Lenny Rachitsky @lennysan quoted @embirico saying coding agents are critical for building super assistants that can perform tasks autonomously.
PMs as capital allocators: Claire Vo @clairevo noted PMs often act as de facto capital allocators, deciding where scarce engineering resources drive business or technical impact.
AI Industry Developments & News
Automated feedback for theorists: Google Research @GoogleResearch announced an experiment using Gemini 2.5 Deep Think to provide automated feedback on theoretical CS papers, with 97% of STOC2026 authors finding it helpful.
AI safety research partnership: Google DeepMind @GoogleDeepMind announced an expansion of its AI safety and security research partnership with the AISecurityInst, focusing on model monitoring and social impact analysis.
From LinkedIn • Deeper Insights
Product Management Insights & Strategies
Start Simple with Memory Layers: Udi Menkes breaks down how ChatGPT’s long-term memory really works: four targeted layers handling session context, user-consented facts, recent summaries, and the sliding window of active dialogue. Each layer solves a clear problem—adaptation, retention, continuity, coherence—showing PMs how to design personalization without needless complexity.
Building Agility for Live Pivots: Marc Baselga outlines why candidates stumble in Uber’s 60-minute “Jam” stress test: they rehearse rigid scripts instead of a flexible skeleton. His drills—interrupt-driven practice, pivot-statements, curveball recordings—train PMs to triage new constraints, ask clarifying questions, and decide on-the-fly, mirroring real-world product chaos.
AI Tools & Applications
Match Tools to Your Development Stage: Paweł Huryn compares prototyping studios—from low-friction drag-and-drop (Lovable, Dyad) to code-centric frameworks (Firebase Studio, Cursor). He advises PMs to start with the simplest tooling that fits team skills and project maturity, avoiding premature engineering overhead until you’re ready for production workflows.
Evaluating Models with Parallel Agents: Tal Raviv walked through a personal dilemma using Cursor’s multi-agent feature, feeding the same context to Gemini 3, Opus 4.5, and GPT 5.1 to compare “thinking styles.” His step-by-step workflow shows PMs how to set up parallel chats, tag diverse inputs, and surface which AI approach best matches your decision-making needs.
AI Product Launches & Updates
Figma in ChatGPT Goes Beyond Diagrams: Yuhki Yamashita announces the Figma App for ChatGPT now converts any conversation into Figma Slides decks or set of branded marketing assets. This end-to-end chat-to-design flow accelerates ideation and hands off seamlessly to Figma for team editing and collaboration.
From YouTube
Sahil Bloom Gives You a Plan for 2026
Greg Isenberg • December 15, 2025
In this episode, Sahil Bloom and Greg Isenberg guide viewers through a seven-question personal annual review—covering mindset shifts, energy creators and drains, limiting “boat anchors,” fears, and key wins and misses—to mine insights from 2025 and craft a clear plan for 2026.
Key Takeaways:
- “What did I change my mind on this year?” prompts using your January 2025 calendar to relive past mindsets and spot true perspective shifts by noting what now makes you cringe.
- Divide activities into professional, personal, and people buckets and review your full-year calendar to mark tasks that created energy versus those that drained it, because outcomes follow energy.
- “Boat anchors,” defined as mindsets, behaviors, or relationships that drag performance, should be identified and “cut the line” as the fastest way to free up capacity for growth rather than adding new routines.
NVIDIA Nemotron 3 Nano 30B First Impression - Shipmas Day 11
All About AI • December 15, 2025
All About AI tests NVIDIA’s new Nemotron 3 Nano 30B A3B mixture-of-experts model—demonstrating its rapid API performance, 1 million-token context window, and open-source weights by building and running Python demos in open code.
Key Takeaways:
- Nemotron 3 Nano 30B A3B is a 30 billion-parameter mixture-of-experts model with only 3 billion parameters active at a time, delivering 4Ă— the throughput and 60% fewer reasoning tokens than its predecessor.
- The model offers a 1 million-token context window, fully open weights with shared data sources, and is available today on Hugging Face and other platforms.
- Live demos via NVIDIA’s API showcase sub-second responses for text-to-image generation, Streamlit UI creation, and complex tool-chain workflows (web fetch, file I/O, plotting) with high accuracy.
How Zapier's EA built an army of AI interns to automate meeting prep and scale internal alignment
How I AI Podcast • December 15, 2025
As Executive Assistant to the CEO at Zapier, Courtney Hickey shows how she built Zapier Agents and custom GPTs to automate weekly meeting preparation, generate AI-driven feedback, and streamline pre-executive document reviews. Her workflows reinforce Zapier’s cultural values, improve decision-making, and scale strategic alignment across the organization.
Key Takeaways:
- Hickey’s weekly meeting prep agent runs every Friday at 8 a.m., scans her calendar, researches external attendees via web search and HubSpot, creates Todoist tasks two hours before each meeting, and posts a structured prep digest to Slack with personalized insights.
- She automates post-meeting coaching by feeding transcripts from tools like Fathom and Fellow into a Zapier Agent that evaluates participants against Zapier’s five dysfunctions of a team, decision-making norms, and company values to deliver two specific growth opportunities and successes to each attendee.
- Courtney built a custom GPT in ChatGPT that embeds Zapier’s strategy memos, operating principles, and Wade’s feedback style to review and sharpen “t-up” strategy docs, suggesting clearer recommendations, trade-offs, and executive-ready rewrites before they reach the CEO.