|
In this newsletter:
- Google releases Lyria 3 Pro, which generates music tracks
- OpenAI shuts down Sora social app six months after launch
- Claude Code launches auto mode with safety approval filters
Plus, you’ll find new AI tools and this week’s top AI news headlines!
🎵 Google Releases Lyria 3 Pro, Which Generates Three-Minute Music Tracks
Google launched Lyria 3 Pro music generation model one month after Lyria 3, extending track length from 30 seconds to three minutes with better creative control and customization.
What's new:
- Available now for paid Gemini app subscribers
- Users can create tracks up to three minutes long vs 30 seconds with Lyria 3
- Specify different elements in prompts like intros, verses, choruses, and bridges
- The model understands the track structure better than its predecessor
- Rolling out to Google Vids video editing app and ProducerAI (GenAI music production tool Google acquired last month)
- Available in enterprise tools via Vertex AI (public preview), Gemini API, and AI Studio
How it works:
- Trained using data from Google's partners, permissible data from YouTube, and Google
- Model doesn't mimic artists but takes "broad inspiration" if the artist is specified in the prompt
- All tracks created with Lyria 3 and Lyria 3 Pro are marked with a SynthID watermark to denote that AI was used
Why it matters:
Three-minute tracks mark the shift from AI music as a novelty to a practical tool for content creators who need background tracks without licensing fees or waiting on composers.
Google's enterprise push through Vertex AI and ProducerAI positions it as the YouTube of music generation - controlling both creation and distribution - while Spotify and Deezer scramble to build detection tools that protect artists from the flood of AI-generated content.
👀 Read more about Google Lyria 3 Pro launch!
📱 OpenAI Shuts Down Sora Social App Six Months After Launch
OpenAI announced it's shutting down Sora, the TikTok-like social app for AI-generated videos that launched six months ago, without providing a reason or discontinuation date.
What happened:
- Sora peaked in November with 3,332,200 downloads across iOS and Google Play
- Declined to 1,128,700 downloads by February
- Made approximately $2.1 million from in-app purchases in its lifetime
- ChatGPT has 900 million weekly active users for comparison
The problems:
- Launched as an invite-only AI-first TikTok with a vertical video feed
- "Cameos" feature lets users scan faces and make realistic deepfakes (renamed to "characters" after Cameo won a lawsuit)
- Under-moderated content included creepy Sam Altman videos, deepfakes of Martin Luther King Jr., and Robin Williams
- Users intentionally made content with copyrighted characters like Mario smoking weed, Naruto ordering Krabby Patties, and Pikachu doing ASMR
- Easy to evade OpenAI's guardrails, preventing videos of public figures who didn't opt in
The Disney deal that collapsed:
- Disney gave OpenAI a $1 billion investment and licensing deal for characters from Disney, Marvel, Pixar, and Star Wars
- No money changed hands before the deal collapsed with the Sora shutdown
- Disney said it would "continue to engage with AI platforms" going forward
What's next:
- Sora 2 video and audio generation model is still available behind the ChatGPT paywall
- Other companies continue making similar technology accessible
Why it matters:
OpenAI's deepfake social network failed because nobody wanted an AI-only feed despite the impressive underlying technology, proving that viral novelty doesn't translate to sustained engagement.
The shutdown kills the Disney licensing deal that could have legitimized AI-generated content featuring major IP, leaving the industry back at square one on the copyright question while the Sora 2 model remains available for anyone willing to pay ChatGPT subscription fees.
👀 Read more about OpenAI shutting down Sora!
🛡️ Claude Code Launches Auto Mode That Approves Safe Actions and Blocks Risky Ones
Anthropic introduced auto mode for Claude Code, a middle-ground permissions system that lets Claude make approval decisions while blocking potentially destructive actions.
What's new:
- Available now as a research preview for Claude Team users, rolling out to Enterprise and API users in the coming days
- Works with Claude Sonnet 4.6 and Opus 4.6
- Classifier reviews each tool call before it runs to check for destructive actions like mass deleting files, sensitive data exfiltration, and malicious code execution
- Safe actions proceed automatically, risky ones get blocked, and redirect Claude to a different approach
- If Claude insists on continually blocked actions, it triggers a permission prompt to the user
- Enable with claude --enable-auto-mode command or toggle in Settings
The problem it solves:
- Default Claude Code permissions ask approval for every file write and bash command
- Can't kick off a large task and walk away since Claude requests frequent human approvals
- Some developers use --dangerously-skip-permissions to bypass checks, but this creates dangerous outcomes outside isolated environments
- Auto mode provides a safer alternative for longer tasks with fewer interruptions
Limitations:
- Reduces risk compared to skipping permissions, but doesn't eliminate it entirely
- Still recommend using in isolated environments
- A classifier may allow some risky actions if the user's intent is ambiguous or Claude lacks environmental context
- May occasionally block benign actions
- Small impact on token consumption, cost, and latency for tool calls
Admin controls:
- Disable for CLI and VS Code extension by setting "disableAutoMode": "disable" in managed settings
- Disabled by default on the Claude desktop app, toggle on in Organization Settings → Claude Code
Why it matters:
Auto mode addresses the core tension in AI coding agents between safety and autonomy, letting developers step away from their keyboard without risking catastrophic file deletions or data leaks.
The classifier-based approach is Anthropic's answer to competitors who let agents run wild, though admitting it still requires isolated environments reveals the fundamental challenge of trusting AI with system-level permissions.
👀 Read more about Claude Code auto mode upgrades!
My Latest LinkedIn & X/Twitter Posts:
- 10 powerful ways to use ChatGPT Deep Research (view post)
- 9 ChatGPT prompts to extract strategies and find insights (view post)
- Prompting cheat sheet for ChatGPT, Perplexity, Grok & Gemini (view post)
- How to utilize powerful Notebook LM (view post)
- 10 ChatGPT prompt templates to get better results faster (view post)
In partnership with Kit:
As a creator, your time should be spent doing what you love, not juggling a dozen tools just to run your business.
Kit gives you everything you need in one place:
✅ Build and grow your email list (I use Kit for my newsletter)
✅ Easily monetize with paid newsletters and digital products
✅ Automate your emails with triggers and custom workflows
✅ Track what’s working and optimize with powerful insights
It’s not just email. It’s your entire creator business, simplified.
Join a thriving community of successful creators.
Use Kit for Free — and Start Building Smarter
New AI Tools to Boost Your Productivity:
- Honestly: Embeds Reddit and TikTok reviews on Shopify product pages.
- DataSieve: Extracts emails and phone numbers from unstructured text.
- NextPhone: AI receptionist that answers calls and books appointments.
- Clip Task: Converts screen recordings into structured project tasks.
- PIO: Manages global hiring, payroll, and compliance.
- Venn: Connects AI models to business apps for automated actions.
- MacNotch: Turns the MacBook notch into a live activity dashboard.
- Fastlane: Turns trending videos into branded content for TikTok, Reels, and Shorts.
- AdsTurbo: Generates AI ad creatives and copy for social media.
- AutoSend: Connects local databases to AI via Model Context Protocol.
- ClipLedger: Track campaigns and calculate creator payouts without manual work.
- ArticleBack: The email platform that an AI agent can operate.
- Pendium: Optimizes brand visibility and rankings across AI search engines.
- SupaPin: Automates Pinterest growth by generating and publishing SEO-optimized pins.
- Fantastical: Smart calendar with natural language event scheduling.
- Claude Cowork: Organizes tasks, files, and context into local, privacy-focused projects.
This Week's Top AI News Headlines:
- Google Gemini Now Lets Users Import Chats and Personal Data From Other AI Chatbots to Make Switching Between Assistants Easier (View Article)
- Apple is Reportedly Planning to let Claude and Other AI Chatbots Integrate With Siri in iOS 27 to Expand Voice Assistant Capabilities (View Article)
- Meta Prepares Third-Generation Ray-Ban Smart Glasses With Enhanced AI Features and Improved Hardware for Everyday Wearable Computing (View Article)
- Wikipedia Tightens Rules on AI-Generated Content, Restricting Use of AI Tools in Article Writing to Protect Accuracy and Editorial Integrity (View Article)
- Google Expands Search Live Globally With AI that Uses Voice and Camera Input to Deliver Real-Time Visual and Spoken Search Results (View Article)
- ByteDance Brings Dreamina Seedance 2.0 AI Video Generation Model to CapCut to Enable Advanced Text-to-Video Creation and Editing (View Article)
- Conntour, an AI Video Search Startup, Raises $7 Million to Build a Search Engine that Analyzes Security Camera Footage Using Computer Vision (View Article)
- Cohere, an Enterprise AI Model Startup, Launches Open-Source Voice Model Designed for High-Accuracy Speech Transcription and Developer Customization (View Article)
- Mistral, a European AI Startup, Releases New Open-Source Speech Generation Model to Compete With Proprietary Voice AI Systems (View Article)
- Manus AI Hype Cycle Enters Predictable Phase as Early Excitement Gives Way to Real-World Limitations and Growing Skepticism (View Article)
- AI Skills Gap Widens as Companies and Power Users Rapidly Advance With AI Tools While Average Workers Struggle to Keep Up (View Article)
- Google Explores “TurboQuant” AI Memory Compression Technique to Reduce Model Costs and Improve Efficiency Inspired by Silicon Valley’s “Pied Piper” Concept (View Article)
- Reddit Introduces New Human Verification Requirements to Crack Down on Bots and Ensure Authentic User Activity on the Platform (View Article)
- Webtoon Launches New Platform With AI Translation Tools and Revised Revenue Model to Help Creators Reach Global Audiences and Earn More (View Article)
Work With Me:
If you enjoy this newsletter, please forward it to your friends and colleagues.
Follow me on LinkedIn and X/Twitter to see my latest posts.
Have a wonderful week!
Andrew Bolis
|