Veo 3 Now GA in Gemini API
Today's curated insights on AI product management from X/Twitter across 60+ expert sources and YouTube channels.
Veo 3 Now GA in Gemini API
From X
AI Product Launches & Updates
All-in-one speech recognition model: Alibaba Qwen @Alibaba_Qwen introduced Qwen3-ASR with high-accuracy EN/CN plus nine additional languages, auto language detection, <8% WER, and robust performance in noisy, low-quality, and far-field conditions.
Veo 3 GA in Gemini API: Sundar Pichai @sundarpichai announced Veo 3 and Veo 3 Fast are now GA in the Gemini API, adding support for 9:16 vertical format, 1080p HD output, and a nearly 50% price reduction.
Perplexity for Government: Arav Srinivas @AravSrinivas launched a zero-data-usage, fully secure AI service for government users, offering access to premium models without logins, enterprise contracts, or data logging.
AI Tools & Applications
Open-source vibe coding platform: Guillermo Rauch @rauchg shipped an OSS “vibe-coding platform” built with the Vercel AI SDK, Gateway, and Sandbox, featuring a tuned GPT-5 agent loop that can read/write files, run commands, install packages, and autofix errors (demo: one-shot multiplayer Pong in Go).
Design systems in v0: v0 Team @v0 introduced Design Mode for generating unique color schemes, custom brand systems, and light/dark previews, with upcoming support for fonts, custom components, and AI-generated designs.
Product Management Insights & Strategies
Unfiltered PM tool review: George Nurijanian @nurijanian shared his weekend testing of 20+ PM tools, noting that while Jira remains indispensable, most AI PM tools focus on PRD automation rather than solving core workflow challenges.
AI agent production challenges: Pawel Huryn @PawelHuryn warned that AI agents are “brittle in production” and emphasized that successful implementations are essentially LLM workflows, where orchestration outweighs autonomy for cost, latency, and reliability.
AI Industry Developments & News
AI governance endorsement: Anthropic AI @AnthropicAI endorsed California State Senator Scott Wiener’s SB 53, advocating a regulatory approach based on transparency rather than technical micromanagement.
Agentic AI ROI survey: Phil Schmid @_philschmid highlighted a Google Cloud survey of 3,466 global enterprise leaders showing 88% of agentic AI early adopters report positive ROI and 74% see Gen AI ROI within one year.
From YouTube
Screensharing How This Guy Makes Millions with AI Agents
Greg Isenberg • September 08, 2025
Greg Isenberg interviews Rowan Cheung of The Rundown to unveil seven AI-driven workflows—from AI avatars and GPT co-editors to scheduling and onboarding agents—that let his 15-person team operate at a 50-person scale and generate millions in revenue.
Key Takeaways:
- The Rundown uses HeyGen for video avatars and 11Labs for voice to transform newsletter stories into short-form clips, growing an AI-driven Instagram following to over 150,000 in under a year.
- They built a Lindy-powered agent named “Ava” that auto-reads Gmail threads, checks Google Calendar slots, and books or reschedules meetings instantly by CC’ing Ava’s address.
- A custom GPT assistant trained on SOPs, Loom transcripts, HR policies and team roles answers new hires’ questions around the clock, streamlining onboarding and reducing manager load.
How Devin replaces your junior engineers with infinite AI interns that never sleep | Scott Wu (CEO)
How I AI Podcast • September 08, 2025
In this How I AI Podcast episode, Scott Wu, CEO of Cognition Labs, demonstrates how Devin, his AI agent, acts as an asynchronous junior engineer by scoping tasks, using Deep Wiki to navigate code, and autonomously generating GitHub pull requests for features like a chat PRD MCP server integration, front-end fixes, documentation, testing, and incident triage.
Key Takeaways:
- Devin leverages Deep Wiki to AI-generate repo documentation, locate specific files or components like the MCP server list, and provide combined natural-language explanations with code snippets.
- Wu refines prompts by embedding contextual code details—such as type definitions and function names—before launching async Devin sessions that produce ready-to-merge pull requests without real-time oversight.
- Cognition Labs tags Devin first in Slack for incident triage, enabling it to fetch crash reports from Sentry, trace recent changes, draft diagnostic reports, and run local tests to accelerate on-call workflows.