Google launches new offline model for your phone

Hola Decoder😎

If someone forwarded this to you and you want to Decode the power of AI and be limitless, then subscribe now and Join Decode alongside 30k+ code-breakers untangling AI.

🍌 Google Drops Two New AI Models for On-Device Search and Image Creation

Google introduced EmbeddingGemma 2 for searching across text, code, images, video and audio locally, alongside Nano Banana 2.1, its upgraded image generation and editing model.

Everything Becomes Searchable - The 740M-parameter EmbeddingGemma 2 puts text, code, images, video and audio into one shared space, enabling searches like finding a specific video moment from a voice memo.

AI Search Stays On-Device - The model can run completely locally and offline, using as little as ~567MB active RAM for multimodal weights, keeping private photos, recordings and documents off the cloud.

Nano Banana Gets an Upgrade - Nano Banana 2.1 improves visual design, mask-based editing and subject consistency, aiming for more natural-looking generations and more precise edits.

Rolling Out Across Google - Nano Banana 2.1 is rolling out across Gemini, AI Mode in Search, AI Studio, Flow, Stitch, Google Ads and Gemini Enterprise Platform.

Google is pushing AI creation and understanding in two directions at once. EmbeddingGemma 2 brings multimodal retrieval onto local devices, while Nano Banana 2.1 expands higher-quality image creation and editing across Google’s consumer, developer and advertising products.

Together with Tatari

Bring Your Toughest Growth Question to Forward 

Bring the acquisition question your team has been circling for months. 

On October 29, Forward by Tatari gives you direct access to leaders making growth decisions at scale.

Inside The Glasshouse in NYC, you can:

• Ask Reddit CEO Steve Huffman how discovery is changing across a platform reaching 127 million people daily.
• Hear how Cody Plofker helped scale Jones Road Beauty to nine figures and what he would prioritize now.
• Learn how Liquid I.V., MANSCAPED, Thorne, and Avocado Green Brands make TV accountable to revenue.

This is a working afternoon for marketers deciding where to put their next dollar. You’ll see how brands plan major TV moments, connect reach to sales, and broaden acquisition without leaving Meta.

Forward is free to attend, but seats inside The Glasshouse are limited.

Secure your spot and bring back a growth plan your team can use.

🌐 How to Build a Real Hosted Website From One Prompt Using ChatGPT Sites

One description. One message. A live website at a shareable link. No templates, no hosting bills, no weekend wasted. Here's how.

Step 1: Switch to Work mode Go to chatgpt.com and open the Work tab. Sites lives here, not in a normal chat. Requires ChatGPT Plus or Pro.

Step 2: Describe the site with your real content Attach your logo and photos, then paste this:

"Build a website for [BUSINESS NAME], a [WHAT IT IS] in [CITY]. Audience: local customers, mostly on phones. Pages: Home, Services with prices, About, Contact. Content: [PASTE YOUR SERVICES, PRICES, AND ABOUT PARAGRAPH]. Contact: tap-to-call button for [PHONE] and a Book button linking to [YOUR BOOKING LINK]. Do not invent reviews, prices, or facts. Wherever I have not given you content, put [PLACEHOLDER] so I can see it. Mobile first."

The "do not invent" line matters. Without it, ChatGPT fills every gap with convincing fake content.

Step 3: Edit by talking No drag-and-drop. Every change is a sentence. Be specific and do one or two edits at a time:

"Make the homepage headline [YOUR HEADLINE] and move the Book button directly under it. Show prices as a table on the Services page. Keep everything else exactly as it is."

Step 4: Publish and share When the preview looks right, select Share > Anyone on the internet > Publish. Copy the link. Your site is live at a chatgpt.site address. To connect your own domain, go to site settings > Add domain and paste the DNS records into your domain provider.

Takes under an hour. Already included in your ChatGPT plan.

One thing it will not do: take payments. For a shop, link out to Stripe, Shopify, or Gumroad.

⚡ Anthropic Launches Its Fastest and Cheapest Claude Yet

Anthropic launched Claude Haiku 5.5, a small model built for high-volume tasks like summarization, classification, customer support, browser use and subagent work, with major improvements in both price and performance.

Haiku Gets Much Cheaper - Haiku 5.5 costs around 75% less per task on average than Haiku 4.5, with requests under 100K tokens priced at $0.10/M input and $0.50/M output.

Small Model, Big Performance Jump - Haiku 5.5 scored 72.4% on OSWorld, up from 15.7% for Haiku 4.5, while beating GPT-6 Luna on several supplied coding, knowledge-work and computer-use benchmarks.

Speed Becomes the Selling Point - Anthropic calls Haiku 5.5 its fastest standard-speed model yet, targeting live support, browser tasks and repetitive agent work. Asana reported up to 2.5x faster inference per agent turn.

Sonnet Gets Cheaper Too - Anthropic also cut Sonnet 5.5 cache-read pricing by 50%, reducing typical agentic-task costs by around 20%, while Max and Team subscribers are getting monthly API credits.

Haiku 5.5 is not meant to replace Sonnet or Opus on the hardest agentic coding jobs. Anthropic is positioning it as the fast, inexpensive worker for narrower tasks, including subagents, while larger Claude models handle the difficult reasoning around them.

Together with Playbook Pro

Your Next 1,000 Customers Are Already Here

Pouring more money into ads won’t fix flat growth. The real leverage is hiding inside the customers you already have, and every day you wait, that momentum leaks away.

Networked Growth Engines is the 26-page field guide built from 200+ hours of research and real brand wins. It gives you ready-to-use systems you can plug in this week, no giant team or budget required. With it, you can:

🔥 Spark self-propelling communities using the Community Building Playbook, exactly how Nyfter built a 500,000-member Discord that drives product innovation.

📈 Design referral programs that fund themselves with the Referral ROI Calculator, the same loop that powered 34% of Blue Apron’s customer acquisition and delivered a 17.3× ROI for Latico Leathers.

🤝 Form trust-based partnerships fast using the Partnership MoU and 50-template Outreach Library to unlock new audiences without paid ads.

Your current customers are the growth engine. 

Download the guide today and start compounding growth before another week of ad spend vanishes.

🏆 Tools you Cannot Miss:

📱 Ava – An AI video editor that analyzes raw footage, assembles cuts from creative directions, revises edits from feedback, and produces different hooks, scripts, and aspect ratios. 

📸 MacCam – Native Mac camera app for shooting photos and videos from a Mac, iPhone, or webcam with portrait lighting, photographic styles, and direct Apple Photos saving. 

🗜️ zipo.pics – Compresses PNG, JPG, WebP, GIF, and AVIF images directly in the browser without uploading them, including batch processing and format conversion. 

🖼️ PicoAsset – Converts and optimizes batches of images and videos directly from Finder, including WebP, AVIF, HEIC, MP4, MOV, WebM, and MKV formats. 

🖥️ Screenbits – Records tutorials, product demos, gameplay, webcam, microphone, and system audio in up to 4K with live annotations, cursor effects, overlays, and zooms

🥳 Prompt of the day

Create a Photorealistic Yor Forger Mirror Selfie

Image Prompt: Create an ultra-photorealistic vertical iPhone 16 Pro Max flash mirror selfie of Yor Forger inside a Japanese bamboo forest-inspired nighttime fitting room, combining authentic smartphone photography with elegant ceremonial styling.

Context

• Character: Yor Forger
• Setting: Bamboo Forest Night fitting room
• Outfit: Deep crimson floral yukata with black obi
• Hair: Long black hair styled in her elegant signature updo
• Footwear: Lacquered geta
• Accessories: Braided crimson shimenawa-style cord, tiny gold bells, blank sealing charms, ornamental ritual wrist cuff
• Prop: Harmless prop kunai
• Photography: iPhone 16 Pro Max mirror selfie
• Aspect ratio: Vertical 9:16

Instructions

Show Yor taking a natural mirror selfie with her face completely visible and recognizable, maintaining a calm, perfectly composed expression. Dress her in a richly textured deep-crimson floral yukata secured with a black obi, paired with lacquered geta and subtle ceremonial accessories including a braided crimson shimenawa-inspired cord, tiny gold bells, blank paper sealing charms, and one decorative ritual wrist cuff. Have her hold a clearly decorative prop kunai pointed downward in a relaxed, non-threatening manner. Surround her with dense bamboo, softly glowing stone lanterns, a distant torii gate, scattered maple leaves, and faint violet atmospheric fog. Capture everything with harsh direct smartphone flash, realistic reflections, natural exposure falloff, slight computational sharpness, and believable silk and fabric textures.

🤳AI Nugget of the Day

Thanks for Decoding with us🥳

Your feedback is the key to our code! Help us elevate your Decode experience by hitting reply and sharing your input on our content and style.

Keep deciphering the AI enigma, and we'll be back with more coded mysteries unraveled just for you!