
Good Morning! Here's what I have for you in today's newsletter:
Google launches Gemini 4 Neon, its mid-tier model at 3/15 per million tokens
OpenAI adds Live Dictation to ChatGPT, your words appear as you talk
Netlify lets you describe an app and have agents build it
Turn raw clips into Reels with Claude and Instagram Edits for Mac
4 new AI tools worth trying today
AI MODELS
Google released Gemini 4 Neon for coding, knowledge work, and science. Positioned between Carbon and Hydrogen, it has a 1 million token context window and costs $3 per million input tokens and $15 per million output tokens.
Google says Neon ranks first on the Artificial Analysis Coding Agent Index and third on its Intelligence Index. For debugging and refactoring, it can make targeted changes across multiple files.
At the default thinking level, Google reports a 32% lower cost per task than Carbon. Cached input costs $0.15 per million tokens, a 95% discount.
Four thinking levels let developers adjust the balance between speed and depth. Neon is available to AI Pro and Ultra subscribers in Gemini and Antigravity, through the paid Gemini API, and in Gemini for business.

Neon follows yesterday’s GPT-6.2 Sol and Microsoft MAI releases, extending a week focused on capable models at lower prices. Google’s reported coding lead gives that competition another dimension: mid-tier models are challenging flagships on performance as well as cost. For developers, the choice increasingly depends on the cost of a completed task, rather than the model’s size or tier.
AI TOOLS
OpenAI rolled out Live Dictation in ChatGPT, displaying speech as text while users talk. It’s rolling out globally across all plans, including Free, on iOS, Android, web, and desktop.
Transcribed words appear directly in the message box during recording, removing the wait to see the text after a long message.
Tap or click the waveform to show or hide the transcript while speaking. Users can follow the text or leave it out of view.
The previous 10-minute recording limit has been removed, allowing longer briefs, meeting recaps, and rough drafts in one recording.

Recent voice launches, including Muse’s live conversations and Gemini’s Voice Remixing, have focused on spoken interaction and generated speech. OpenAI’s update improves a more basic step: getting instructions into an assistant. As agents take on longer assignments, visible transcription gives people a way to review a detailed spoken brief before putting it to work.
AI AGENTS
Netlify unveiled a redesigned platform where agents build apps from descriptions, with changes visible in a live preview. Most new users have access now, while existing accounts are receiving it gradually.
Agents can use more than 150 connectors, including Notion, Stripe, Airtable, and Linear. These run on credit-based plans, and deployed apps can continue using the connected services.
Business users can request an app or a change, then developers can continue the same project through Git, an IDE, or a coding assistant. Teammates review changes before deployment.
Netlify Server supports chat apps, agents, and longer-running tasks. Each deploy preview gets a dedicated server for testing. Netlify says more than 10,000 servers already run on it.

Last week, Cursor added live app building and Automattic introduced Spacefast for publishing agent-built projects. Netlify now brings generation, business connectors, and deployment into its existing platform. Together, these releases show the competition moving beyond generating a prototype toward hosting, maintaining, and extending the application after it ships.
HOW TO AI
Claude builds the cut. You check the captions and post.
Instagram’s free Edits for Mac app lets Claude cut clips, add captions and build a Reel in its timeline. You’ll need a Mac with Apple silicon (M1 or later), an Instagram account, Edits for Mac and Claude Code.

Step 1: Get Edits ready
Download Edits for Mac and sign in with Instagram. Create content-library inside Movies, with raw and stats folders inside it. Put one talking-head clip in raw. Edits can’t open files in Downloads or Desktop. Download Edits
Step 2: Connect Claude Code
In Edits, open Settings, then Automation, and turn on Enable Automation Server. Click Copy next to Claude Code and run the command in Terminal. In Claude Code, type /mcp and look for a green check beside metapilot-native. Create a project in Edits and leave it open so Claude can access the editing tools.

Step 3: Plan from successful Reels
Screenshot the openings of three popular Reels on your topic with their captions visible, or use your best Reels’ stats. Save them in stats. Ask Claude to compare them, transcribe your clip and write a storyboard under 45 seconds. Transcription places the clip on the timeline, so tell Claude to wait for approval before cutting.
Step 4: Let Claude edit
Ask Claude to build the first 5 seconds with the hook, title and captions. Check the font, caption position and pacing. Then approve the full cut, asking it to remove silences and filler words and set the project to 9:16. Give precise notes, such as “move the cut at 0:12 half a second later”.

Step 5: Check and publish
Play the Reel with sound, then muted. Correct misheard words in the Captions panel’s Edit tab. Click Export in Edits yourself, since exports started by Claude can lose the music track. Select Share to Instagram, add your caption, location and audience, then post.
P.S. Become an AI Operator here.

Google rolled out Genie 4 in Project Genie for eligible AI Ultra subscribers aged 18 and older, with an AI Pro waitlist. Users can change worlds through prompts during a session, hear generated audio, and remix creations from a community gallery.
Tencent released Hy4 Flash weights on Hugging Face under the Tencent Hy License, alongside FP8 and Base versions. The model has 220 billion total parameters, 12 billion active, and a 1 million token context window. Running it requires multiple GPUs.
Meta added virtual try-on to Muse. During clothing searches, the shopping agent can offer a preview of how an item might look on the buyer.

🎵 Suno: Opened its MCP connector to all plans, including Free, for creating, covering, remixing, and extending songs through ChatGPT, Claude, Gemini, and Cursor.
🐛 CodeRabbit: Launched Verify in free beta. The agent runs apps in a sandbox, tests pull request changes, and documents bugs with screenshots.
⌨️ Raycast: Added a local MCP server that gives Claude, Codex, and Cursor access to notes, clipboard history, snippets, browser tabs, and settings.
🤖 Vercel Board: An experimental message board for coordinating work between teammates and agents, with threaded tasks and two-way Slack and GitHub sync.
THAT’S IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
✓ Full archive of premium guides with ready-to-use prompts
✓ Structured AI courses (step-by-step, start-to-finish)
✓ Every upcoming premium tutorial
✓ $500+ worth of AI tools




