- AI PlanetX
- Posts
- Google Unveils Gemini Embedding 2
In partnership with
Welcome to another edition of AI PlanetX.
Google unveils Gemini Embedding 2; ChatGPT adds interactive visuals for math and science; Gemini expands in Chrome to India, Canada, and New Zealand.
Inside This Edition: 💎
Hottest AI News
Top AI & SaaS Tools
AI Tutorial: Turn Reddit Into Viral LinkedIn Post
Top AI & Tech News
AI Art Spotlight
Prompt of the Day: McKinsey Market Research Prompt
Featured AI Video
Course of the Day: Gemini CLI - Code and Create
Hottest AI News
DeepMind
Google Unveils Latest Multimodal Gemini Embedding 2 Model
Gemini Embedding 2 has officially launched in public preview, giving developers, researchers, and enterprises worldwide the ability to use it via the Gemini API and Vertex AI to build advanced AI applications.
Details:
The model runs in a unified embedding space supporting text, images, video, audio, and documents (up to six pages), helping systems connect different data types across 100+ languages
It handles up to 8,192 text tokens, videos up to 120 seconds, and native audio, while using Matryoshka Representation Learning with output sizes of 3072, 1536, or 768
Building on Google’s multimodal research, it already outperforms leading benchmarks in semantic search, data clustering, and Retrieval-Augmented Generation
With this release, Google sets a higher standard for multimodal AI, available through the Gemini API and Vertex AI.
Speak naturally. Send without fixing.
Wispr Flow turns your voice into clean, professional text you can send the moment you stop talking. Not rough transcription you have to clean up. Actual polished text - ready for email, Slack, or any app.
Reid Hoffman sends 89% of his messages with zero edits using Flow. Millions of people worldwide have made it part of how they work, including teams at OpenAI, Vercel, and Clay.
Speak the way you think. Go on tangents. Change your mind mid-sentence. Flow strips the filler, fixes the grammar, and gives you text that reads like you spent five minutes writing it.
Works on Mac, Windows, iPhone, and now Android - free and unlimited on Android during launch.
OpenAI
ChatGPT Launches Interactive Visuals for Math and Science
OpenAI has rolled out an update to ChatGPT introducing interactive visual explanations for math and science, replacing static text with dynamic tools to help users better understand complex concepts.
Details:
The feature covers over 70 core concepts tailored for high school and college-level curricula, including fundamental principles such as the Pythagorean theorem, Ohm’s law, kinetic energy, and compound interest
Users can interact directly with these lessons by adjusting variables and manipulating formulas via sliders, allowing them to watch graphs, diagrams, and geometric renderings update in real-time
By moving beyond simple conversational answers, this update acts as a hands-on laboratory experience that is available to all logged-in ChatGPT users across all plans
This initiative represents a significant shift in AI-assisted education, moving toward more active, experimental learning.
Top AI & SaaS Tools
GSpeech (Life-time Deal): A real-time AI text-to-speech and translation tool that converts website text into lifelike audio
Krisp Mobile: An outgoing call recorder for the Krisp mobile app that auto‑records calls, transcribes them, and creates AI notes [F-R-E-E]
Qwen3-TTS: SOTA 0.6B and 1.7B speech models supporting 10 languages with prompt-based voice design, 3s zero-shot cloning, and 97ms streaming [F-R-E-E]
Kimi Slides: Slide-generation tool integrates with Nano Banana Pro for agentic search, file-to-slide conversion with PPTX export, and designer-quality visuals [F-R-E-E]
Tasklet: AI automation that lets you describe workflows in plain English and then runs those tasks automatically across your tools and apps [F-R-E-E]
Free email without sacrificing your privacy
Gmail tracks you. Proton doesn’t. Get private email that puts your data — and your privacy — first.
AI Tutorial
How to Turn Reddit Threads Into Viral LinkedIn Posts
Most LinkedIn posts fail because they’re written from guesswork. This tutorial shows how to pull real audience language from Reddit and turn it into high-performing LinkedIn content using a simple 7-step workflow.
Pick right thread
Go to a subreddit in your niche. Find a post with strong engagement. Look for lots of comments, not just upvotes.
Trigger JSON view
Enter the thread. Tap the URL bar. Add /.json at the end. Press enter.
Copy raw audience voice
You will see a wall of raw text. Copy everything on the page. This is the unfiltered language of your target audience.
Clean input
Before you paste into an AI model:
Remove obvious repeats if you notice them.
Keep the title, top comments, and replies.
Do not rewrite anything.
Extract frustrations and hooks
Paste the JSON into your AI model. Use this prompt:
"I am pasting raw JSON data from a Reddit thread about [TOPIC]. Identify the top 3 frustrations. Include the exact phrases people used. Write 5 LinkedIn hooks based on those frustrations. Make them short and punchy like these:Example A: Stop doing [X]. It’s why you’re stuck.Example B: Most people think [Y]. That’s why they lose.Example C: If you struggle with [Z], read this.JSON Data:[PASTE DATA]"Turn insights into posts
Move the output to Claude. Use this prompt:
"You are an expert LinkedIn ghostwriter. Write 5 LinkedIn posts using the frustrations below. Rules: Short sentences. Lots of white space. One actionable tip + one framework per post.End with: 'Repost if this helped.'Frustrations and exact phrases:[PASTE]"Final polish before posting
Pick the strongest post. Tighten the first 2 lines. Remove extra words. Post, then save best hooks for later.
Note: Focus on threads with strong comment depth, not just high upvotes. Long discussions usually mean stronger emotional signals.
Top AI & Tech News
Google is expanding its Gemini integration for Chrome to India, Canada, and New Zealand
WordPress launched my WordPres net, a browser-based private workspace that runs the full WordPress software locally in your browser
Elon Musk announced a joint project between Tesla and his AI startup xAI called "Macrohard" (aka "Digital Optimus"), a system designed to emulate functions of software companies
Meta unveiled four MTIA chips (300 in production; 400/450/500 arriving through 2027) built with Broadcom on RISC‑V and TSMC to power AI
AI Art Spotlight
Create Your Own Photo Like This Using the Prompt Below:
A realistic, highly detailed geographic map viewed from above, accurately representing the region where [COUNTRY] is located. Rising from the map is a miniature 3D diorama of [CITY], precisely positioned at its real geographic location. The diorama features iconic architecture, natural landmarks, and cultural elements seamlessly emerging from the map surface. Ultra-realistic materials, fine textures, and depth of field. Soft studio lighting with subtle shadows, cinematic composition, premium macro photography style. Square 1:1 format, high detail, realistic scale.Model: Nano Banana 2
Prompt of the Day
McKinsey‑Level Market Research Prompts
Featured AI Video
I’m changing how I invest my money because of AI.
AI Course of the Day
Gemini CLI: Code and Create with an Open-Source Agent
This course shows how to use Gemini CLI’s agent with Model Context Protocol (MCP) and extensions to run smooth, multi‑tool workflows that connect local files with cloud services such as Google Workspace and Canva. You will learn to automate common software development tasks, including building website features, creating data dashboards, and setting up reliable code review workflows with GitHub Actions.