|
In this newsletter:
- ChatGPT now works inside your browser and completes tasks for you
- Black Forest Labs launches FLUX 3 Klein for 4K generation and precise editing
- Google launches Universal Cart to track holiday shopping and price drops
Plus, you’ll find new AI tools and this week’s top AI news headlines!
🌐 OpenAI Upgrades ChatGPT for Chrome to Compare Tabs and Complete Tasks Across Websites
OpenAI’s new ChatGPT browser extension lets ChatGPT read and work on the sites you’re signed in to, including Gmail and internal company tools.
What’s new:
- Works in Chrome, Edge, Brave, Opera, and Vivaldi through the ChatGPT desktop app
- Side chat opens ChatGPT beside your page in every one of them except Opera, where tasks start in the desktop app
- The side chat stays with the tab where you opened it
- Chats started there appear in the ChatGPT app, and recent chats open in side chat
- Mention any open tab to give ChatGPT that page as context
- Highlight text to ask about one passage without copying the whole page
- Right-click a page and select Ask ChatGPT to open side chat with that page’s context
- Ask about a YouTube video, and ChatGPT uses its timestamped transcript when captions are available
- Start a task by mentioning your browser in a chat, like asking Edge to update a Salesforce account from call notes
- The built-in browser, reached with @Browser, supports sign-in and keeps browsing inside ChatGPT, away from your regular browser profile
- ChatGPT switches between plugins, your browser, and the built-in browser depending on the task
- Chrome file uploads need “Allow access to file URLs” turned on in the extension settings
- ChatGPT asks before using each new website, and you can allow it once, allow that site, allow all sites, or decline
- Allowed and blocked site lists live in Settings and are shared across the supported browsers
- “Allow for all sites” removes every confirmation, and OpenAI flags it as elevated risk
- Browser history access needs your approval each time, with no always-allow option
- Saved memories are used in browser tasks only if your Memories setting is on
- OpenAI stores browser activity only when it becomes part of the chat, such as text ChatGPT reads, screenshots, or summaries, and your ChatGPT data controls apply to it
Why it matters:
Using ChatGPT for web work used to mean jumping between tabs, copying text over, and pasting answers back. That routine cost a little time every day.
Now that step is gone. Less switching means fewer interruptions and more time for the work itself.
👀 Read more about how to set up ChatGPT in your browser!
🎨 Black Forest Labs Launches FLUX 3 Image Klein for Fast 4K Generation and Precise Editing
Black Forest Labs has released FLUX 3 Image Klein, a compact open model that turns text into images and edits existing ones.
What’s new:
Image quality
- Ranks first in text-to-image, editing, and multi-reference generation in the company’s internal tests, ahead of larger open models and the closed Ideogram V4.5 and FLUX.2 [pro]
- Gains 91 rating points in text-to-image and 70 in editing over FLUX.2 Klein 9B, with 38% fewer parameters
- Learns from images, video, and audio together, so results feel grounded in the real world
Creative control
- Generates natively in 4K with pixel-perfect editing
- Accepts up to 10 reference images at once
- Bounding boxes let you place elements or limit edits to one area
Size and speed
- Has 5.6 billion parameters, the smallest in its comparison with rivals at 7 and 12.9 billion
- Includes compressed FP8, INT8, and NVFP4 versions built with NVIDIA’s support
- Takes 11.2 GB of space, or 5.6 GB in the FP8 version
- Runs the INT8 version on NVIDIA RTX graphics cards with as little as 8GB of memory, through ComfyUI or Diffusers
- Creates a 1MP image in about 4 seconds on an RTX 5090 and under 9 seconds on an RTX 4070
- Creates a 1MP image in under 0.6 seconds on an NVIDIA GB300 chip, and a custom-trained version adds around 5% to that time
Customization
- Comes as a base model for fine-tuning and a distilled model built for low latency
- Supports custom training (LoRA) from day one through Ostris’s AI Toolkit and NVIDIA NeMo AutoModel recipes
- Works from day one with ComfyUI, Hugging Face Diffusers, TensorRT-LLM, SGLang Diffusion, and vLLM-Omni
Access and pricing
- Offers the distilled model through the company’s API as flux-3-image-klein, with prompt upsampling and pixel restoration that keeps unedited areas exactly as they were
- Delivers lower API latency than Qwen-Image 2.1, FLUX.2 [pro], Ideogram V4.5, and Krea 2, at 4.54 seconds for text-to-image and 7.18 seconds for editing
- Starts API pricing at $0.009 per 1K images
- Provides base and distilled weights on Hugging Face, plus reference code with a Gradio demo for bounding-box control
Licensing and safety
- Allows free commercial use for individuals and companies earning under $5M a year, along with research and non-commercial use
- Offers larger organizations self-hosted licensing or the API
- Was tested for misuse by the company and the outside safety firm Cinder, and had substantially fewer weaknesses than other popular open models
Why it matters:
Better image quality has usually meant a bigger model and heavier hardware. Klein breaks that pattern by leading its rankings at a size a home computer can handle.
That puts strong image editing on laptops and desktops people already own, keeps private images off other companies’ servers, and makes the model cheaper for teams to adapt. It’s the second open release in the FLUX 3 family, and more are coming.
👀 Read more about FLUX 3 Image Klein launch!
🛒 Google Launches Universal Cart to Track Holiday Shopping and Price Drops
Google has launched Universal Cart in the U.S., a shopping hub that gathers items from different stores in one place. It was previewed earlier this year and arrives as holiday shopping picks up.
What’s new:
- Saves products from different retailers in a single cart, wherever you shop on Google
- Keeps your progress across Search and the Gemini app, so a handbag saved in AI Mode is still there when you add a sweater later
- Watches your whole list in the background using Google’s database of more than 60 billion product listings
- Sends a notification when an item drops in price or comes back in stock
- Understands the perks on the cards in your Google Wallet and helps you choose which one to pay with
- Shows relevant promo codes right in the cart
- Lets you pay with Google Pay in a few taps, or move your items to the store’s own cart and finish on their site
- Works with merchants including Nike, Sephora, Target, Ulta Beauty, Walmart, Wayfair, and many Shopify stores
- Keeps the store as the seller, so confirmations and customer support come directly from them
- Powers checkout with Google’s Universal Commerce Protocol
Why it matters:
Shopping across several stores usually means juggling open tabs, screenshots, and a half-remembered deal. A cart that remembers it all for you takes the scramble out of holiday shopping.
With the busiest season of the year starting, it gives shoppers a way to buy with more confidence and less second-guessing.
👀 Read more about Google’s Universal Cart!
My Latest LinkedIn & X/Twitter Posts:
- 9 powerful ChatGPT prompts for email writing (view post)
- 10 soft skills that lead to bigger career opportunities (view post)
- 9 ways to use AI for work and productivity (view post)
- How AI really works explained in simple terms (view post)
- 60 AI tools to automate work and business (view post)
- 3 levels of prompting to go from basic to expert-level (view post)
- 15 ChatGPT prompts to help you structure research (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:
- Munch Studio: Repurposes existing content into scheduled social posts.
- Audimate: Assists with tasks inside your existing apps.
- Rushilo: Edits videos through Claude Code or Codex.
- Atticus: Controls Mac apps and tasks using voice commands.
- Superdraft: Finds marketing opportunities and creates email campaigns.
- Afterparty: Turns meeting action items into ready-to-review drafts
- SupoClip: Turns long videos into short clips.
- Remocn Studio: Creates editable videos using coding agents.
- AirJelly: Captures work context and automates follow-ups.
- Takes for Mac: Records and edits talking-head videos.
- DemoScreen: Turns product screenshots into narrated demo videos.
- VideoFlow Studio: Creates launch videos from website URLs.
- Rank Kiwi: Analyzes Instagram and YouTube content performance.
- Revenue Scout: Finds social media leads and drafts replies.
- Retriever AI: Automates browser-based research, outreach, and forms.
- Kai: Manages meetings, emails, tasks, and calendars.
This Week's Top AI News Headlines:
- OpenAI Launches Rosalind Academic Plan, Giving Eligible US University Labs Free Access to AI Tools for Drug Discovery, Genomics, and Life Sciences Research (View Article)
- Anthropic Grants Independent AI Safety Researchers Employee-Level Access to Evaluate Claude Models, Internal Safety Practices, and Risk Management (View Article)
- NVIDIA Launches Nemotron 3.5 Super VL, an Open AI Model that Processes Text, Images, and Video to Help AI Agents Complete Complex, Long-Running Tasks Faster (View Article)
- Amazon's Bee Launches Two Personal AI Devices Starting at $75, Including a Wearable Assistant and a Smartphone Attachment for Capturing Conversations and Organizing Daily Information (View Article)
- OpenAI Expands ChatGPT Sites for Enterprise With US and EU Data Residency, Admin Controls, and Secure Connections to Private Company Services (View Article)
- Google Adds AI Scam Detection to Circle to Search, Letting Android Users Check Suspicious Text Messages and Get Safety Guidance Before Clicking Links (View Article)
- OpenAI Upgrades Codex With Additional Computing Capacity and Automatic Recovery to Reduce Model Overload Errors and Prevent Interrupted Coding Tasks (View Article)
- Meta Launches Muse Spark Open, an Advanced Open-Weight AI Model for Complex Reasoning and Coding that Can Run Locally on Apple's M5 Ultra Mac Studios (View Article)
- Perplexity Adds NVIDIA's Nemotron 3.5 Super VL to its AI Platform, Giving Pro and Max Users Access to an Advanced Model for Complex Visual and Multi-Step Tasks (View Article)
- Microsoft Completes Training MAI-Base-2, its New AI Foundation Model, and Begins Preparing a More Powerful Successor to Strengthen its Independent AI Development (View Article)
- Google Expands Gemini 4 Argon to Antigravity, Bringing Advanced AI Reasoning and Multi-Step Coding Capabilities to its AI Development Platform (View Article)
- Perplexity Open-Sources SPACE, its AI Agent Security Platform that Isolates Tasks in Protected Environments to Prevent Unauthorized Access and Unsafe Behavior (View Article)
- ElevenLabs Launches Instant Soundtracks to Automatically Generate Background Music and Sound Effects that Match the Action and Timing of Videos (View Article)
- Cursor Adds Meta's Muse Wonder AI Model for Complex Coding, Multi-Step Debugging, and Large Software Projects With Up to 1 Million Tokens of Context (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
|