In partnership with

This week in AI, browsers, coding, and personalization are getting a major upgrade. From Perplexity’s free AI assistant browser to Google’s autonomous coding agent and Meta’s new personalization signals, here’s what’s making headlines:

🌐 Perplexity Launches Free Comet Browser
Perplexity AI has released its Comet browser worldwide for free, turning it into a personal assistant for search, tab organization, email drafting, and shopping. A paid Comet Plus tier adds premium content from CNN, Condé Nast, and The Washington Post, with mobile and multitasking features coming soon.

⚙️ Google Jules CLI for AI Coding
Google introduces Jules Tools CLI, letting developers control its AI coding agent Jules directly from the terminal. Running on Gemini 2.5 Pro, Jules can clone repos, write tests, build features, and fix bugs autonomously, while ensuring privacy through secure cloud VMs and GitHub integration.

📱 Meta Uses AI Interactions for Recommendations
Starting December 16, Meta will use user interactions with its AI assistants—whether voice or text—to refine content and ad recommendations on Facebook and Instagram. Sensitive topics remain excluded, and users can manage personalization through Ads Preferences and feed controls.

From free AI browsers and AI-driven coding agents to smarter personalization in social media, these moves show how AI is becoming more deeply embedded in the tools we use every day.

Perplexity AI Launches Free Comet Browser Worldwide

Perplexity AI has made its AI-powered web browser, Comet, available for free worldwide after initially launching it in July for $200 a month to Perplexity Max subscribers. Designed as a personal assistant, Comet can search the web, organize tabs, draft emails, and even help with shopping. The move aims to draw users amid competition from Google, OpenAI, and Anthropic, all of which have recently introduced AI browsers or agents. Perplexity also offers Comet Plus, a paid tier featuring content from trusted publishers such as CNN, Condé Nast, and The Washington Post. The company plans to add a mobile version and a “Background Assistant” feature that can handle multiple tasks simultaneously.

Google Launches Jules CLIGoogle Launches Jules CLI

Google has introduced Jules Tools, a new command-line interface for its asynchronous AI coding agent known as Jules. Jules integrates directly with developers' code repositories, cloning the full codebase into a secure Google Cloud virtual machine to understand the project context and autonomously perform tasks such as writing tests, building features, fixing bugs, and updating dependencies. The new Jules Tools CLI lets developers manage and control these tasks directly from their terminal, enabling integration with existing workflows and automation scripts without leaving their development environment. This approach offers greater developer productivity, flexibility, and visibility into AI-assisted coding activities, marking a shift toward AI as an autonomous collaborator in software development. The system runs on Google's advanced Gemini 2.5 Pro model and maintains user privacy by isolating data within secure cloud VMs. Jules also offers GitHub workflow integration and audio summaries of commits.

Find out why 100K+ engineers read The Code twice a week

Staying behind on tech trends can be a career killer.

But let’s face it, no one has hours to spare every week trying to stay updated.

That’s why over 100,000 engineers at companies like Google, Meta, and Apple read The Code twice a week.

Here’s why it works:

• No fluff, just signal – Learn the most important tech news delivered in just two short emails.

• Supercharge your skills – Get access to top research papers and resources that give you an edge in the industry.

• See the future first – Discover what’s next before it hits the mainstream, so you can lead, not follow.

Join 100,000+ engineers who read The Code to stay ahead of the curve.

Meta Enhances Recommendations with AI Interaction Data

Meta is enhancing its content and ad recommendations on platforms like Facebook and Instagram by using interactions with its generative AI features as a new signal to personalize user experiences. Starting December 16, 2025, Meta will incorporate users' conversations and engagements with its AI—whether via voice or text—into its recommendation systems to show more relevant posts, reels, and ads tailored to users’ interests. Users will receive notifications about this update starting October 7, 2025, and can control their experience through tools like Ads Preferences and feed controls. Sensitive topics discussed with Meta AI, such as religious or political views, will not be used for ad targeting. Additionally, personalization based on AI interactions will depend on account linkages across Meta platforms, ensuring privacy and user control remain prioritized.

Hand Picked Video

In this video, learn how to create a Chrome Extension from scratch using Vibe Code, an AI-assisted coding tool. See how to set up key components like manifest, background scripts, and pop-ups. No prior coding needed, Vibe Code simplifies building by generating working code quickly. Follow along to build an extension that interacts with web pages and APIs, perfect for beginners and pros wanting a fast start in Chrome extension development.

Top AI Products from this week

• Resume Loom - Your dream job deserves more than a generic resume. ResumeLoom uses AI to create tailored, ATS-optimized resumes + portfolios that actually get you hired. 🚀 Launch Special: 50% OFF for first 500 users. Grab your spot before it's gone!

• SpotDB – Give your agents a secure, containerized SQL sandbox with MCP or API access enabled with DuckDB. No infra required, quick spin-up cycle. Push any CSV — query, analyze, and move on.

• LFM2-Audio – LFM2-Audio defines a new class of audio foundation models: lightweight, multimodal, and real-time. By unifying audio understanding and generation in one compact system, it enables conversational AI on devices where speed, privacy, and efficiency matter most.

• DogBird - Professional wholesale real estate software for investors, wholesalers & flippers. Property comps, buyer management, market reports & deal flow automation. Start free trial today!

• Lovelaice - Test prompts across any AI model, evaluate results, and optimize performance in one platform. Compare OpenAI, Claude, Gemini and more side-by-side. Get clear rankings and prompt suggestions. Deploy the best version via API- no tech background required.

• Hey Santa - Hey Santa is an AI-powered experience where kids receive a personalized videos (30+ seconds) straight from Santa. It looks cinematic, feels magical, and takes less than a minute. Parents can add details about their child, and the video is delivered in minutes.

This week in AI

• Ring-1T-preview - Trillion-parameter AI model with strong natural language reasoning, excelling in math and coding challenges, open-sourced for community use and feedback.

• AWS-Powered NBA Stats - NBA Inside the Game powered by AWS delivers AI-driven real-time stats, advanced metrics like Shot Difficulty and Gravity, plus interactive fan tools across broadcasts and apps.

• TX-GAIN: MIT’s AI Supercomputer - MIT's Lincoln Lab unveils TX-GAIN, the most powerful AI supercomputer at a US university, delivering 2 AI exaflops for generative AI, simulations, and advanced research.

• Mochi 1 - Open-source 10B parameter AI model generates high-fidelity, photorealistic videos from text prompts with strong motion fidelity and precise prompt adherence.

• IBM-AMD AI Partnership - IBM and AMD partner with Zyphra to deploy advanced AMD GPU clusters on IBM Cloud, enabling large-scale AI training for frontier multimodal models powering enterprise superagent Maia.

Paper of The Day

AI models excel in abstract reasoning with text and match human accuracy but rely heavily on superficial patterns unlike humans. Their visual reasoning remains weaker, though Python tools improve it. They often generate correct rules but struggle to apply them consistently. This gap underscores architectural limitations needing research to enhance AI’s conceptual understanding, generalization, and explainability. Human-like abstraction and transparent reasoning are crucial for AI to truly generalize and interact effectively with people.

To read the whole paper 👉️ here.