Claude Code Now Available in VS Code

Today’s top AI Highlights:

1. Run 10 AI agents in parallel without creating total chaos

2. Opensource agentic browser that runs AI agents locally

3. ByteDance’s opensource model just changed document OCR forever

4. This $120M AI “cheating” startup got exposed and opensourced on a weekend

5. Claude Code is now available in VS Code

& so much more!

Read time: 3 mins

AI Tutorial

We've been stuck in text-based AI interfaces for too long. Sure, they work, but they're not the most natural way humans communicate. Now, with OpenAI Agents SDK and their text-to-speech models, we can build voice applications without drowning in complexity or code.

In this tutorial, we'll build a Multi-agent Voice RAG system that speaks its answers aloud. We'll create a multi-agent workflow where specialized AI agents handle different parts of the process - one agent focuses on processing documentation content, another optimizes responses for natural speech, and OpenAI's text-to-speech model delivers the answer in a human-like voice.

Our RAG app uses OpenAI Agents SDK to create and orchestrate these agents that handle different stages of the workflow.

We share hands-on tutorials like this every week, designed to help you stay ahead in the world of AI. If you're serious about leveling up your AI skills and staying ahead of the curve, subscribe now and be the first to access our latest tutorials.

Don’t forget to share this newsletter on your social channels and tag Unwind AI (X, LinkedIn, Threads) to support us!

Latest Developments

Run Parallel Coding Agents in Their Own Containers 📦🤖

One AI coding agent refactoring your functions, writing tests, and building endpoints feels like pure magic - the productivity gains feel too good to be true. The natural next step? Let’s run 10 agents in parallel and become a coding superhero.

Chaos ensues! The dream quickly becomes a nightmare of file overwrites, dependency conflicts, and agents that can't coordinate to save their digital lives.

Dagger has built Container Use, an opensource MCP server that gives each coding agent its own isolated containerized environment, complete with Git-backed workflow management. Instead of babysitting one agent at a time or throwing them all into your local environment, you can now run multiple agents in parallel without the chaos.

It works with Claude Code, Cursor, and other MCP-compatible agents, turning your agent workflow from a disaster zone into a controlled, productive operation.

Key Highlights:

1. True Isolation - Each agent gets its own fresh container and Git branch, allowing multiple agents to work on different tasks simultaneously without conflicts or file overwrites.

2. Real-time Monitoring - Track every command and output with detailed logging, giving you a perfect record of what each agent actually accomplished versus what it claimed to do.

3. Emergency Override - Drop directly into any agent's terminal environment to debug stuck processes, inspect file states, and manually steer agents back on track.

4. Git-backed Workflow - All agent work is versioned through Git branches, letting you inspect, merge, or discard agent contributions using familiar Git commands and workflows.

Chrome's Interface Meets Local AI Agents 🌐 🖥️

AI browsers are having their moment, with Opera Neon, Perplexity Comet, Dia, and a dozen other promising to automate your web tasks.

The catch is they all want to run your AI agents on their servers - your browsing data becomes their product.

This small team just forked Chromium's 15-million-line codebase and built Nxtscape, the first fully opensource agentic browser that keeps your AI agents running locally on your machine instead of in some company's cloud.

Nxtscape makes privacy the default, not an afterthought. The browser feels exactly like Chrome but with AI agents that can automate web tasks, semantic search through your history, and even an upcoming LLM-powered ad blocker - all without sending a single byte of your data anywhere.

Key Highlights:

1. True local AI agents - Unlike cloud-based competitors, Nxtscape runs browser-use and computer-use models directly on your machine with full Ollama integration, keeping your browsing history and agent activity completely private.

2. 100% opensource - Built on Chromium with AGPL-3.0 licensing, you can fork the project anytime and avoid the Arc abandonment problem that left users stranded when companies pivot strategies.

3. Familiar yet powerful - Works with all your existing Chrome extensions while adding native AI features like semantic search, ChatGPT-powered bookmarking, and automated form filling without the learning curve.

4. Privacy-first - Designed by developers who aren't in the search or ads business, supporting bring-your-own-keys and local models so your browser never becomes anyone else's product.

Quick Bites

Google’s autonomous AI coding agent Jules now does its homework before touching your code. Jules now automatically scans for AGENTS.md files to grasp your repo's agent ecosystem before making changes, giving it crucial context. The AI also now shows better testing discipline by writing and running tests more reliably on its own.

Mistral has quietly released Small 3.2, a focused update that tackles some practical pain points. The 24B parameter model now handles precise instructions more reliably, cuts down on those annoying repetitive loops that could derail longer conversations, and ships with a more robust function-calling template.

While nothing major, these incremental fixes address real workflow friction points that matter for production deployments.

ByteDance's Dolphin just made traditional OCR pipelines look ancient. Dolphin reads documents like humans do - first scanning the layout to understand structure, then processing each element type (text, tables, formulas) with specialized prompts in parallel. Dolphin is fully opensource with models available on Hugging Face, and includes both page-level parsing for complete documents and element-level parsing for individual components.

We usually keep things technical in this newsletter, but this story perfectly captures how the focus of Silicon Valley money is shifting from innovative technical products to great attention-grabbing machines.

Back in March, two Columbia students built an AI tool to help people cheat on coding interviews. They were suspended, but they turned the backlash into a growth engine. Since then, they’ve gone viral, crafted bold founder personas, and built Cluely, an “undetectable” AI agent that helps users cheat on everything from job interviews and exams to even dates.

The attention strategy worked: $15M Series A from a16z and a $120M valuation.

But just over this weekend, their business got completely exposed in what might be one of the quickest rug pulls in AI history. One developer reverse-engineered the platform, and their desktop app exposed their entire system prompts and models used (mostly GPT 4.1 and Claude 3.7 Sonnet), which says "NEVER mention the specific LLM providers".

On the other hand, another developer then built a complete opensource version called "Horizon Overlay" and released it for free.

Cluely created the new AI playbook - viral content, controversial positioning, blue-chip VC validation. While their code got leaked and cloned, Cluely still sits on $20M+ in funding. This is the new reality: building the product is the easy part; building the story that gets you funded is everything.

Tools of the Trade

1. Deep Graph MCP: MCP server that integrates advanced code graph analysis capabilities into Claude Code for large codebases. It provides advanced code exploration tools like semantic search, dependency analysis, etc., which scale better than Claude's native repository search capabilities.

2. Pipo360: Generate backend code with authentication, database schemas, and CRUD routes across 14+ frameworks and 10+ databases - all from simple language descriptions. It generates production-ready backends in approximately 60 seconds.

3. Taskerio: MCP server for logging and monitoring for AI agents across different projects. It automatically tracks agent progress, captures detailed logs, and sends real-time notifications through push notifications. Works with Cursor, GitHub Copilot, Claude Desktop, and other AI tools.

4. Claude Code for VS Code: VS Code extension that integrates Anthropic's Claude AI directly into the IDE. Comes with features like automatic context sharing of selected text, diff viewing for code changes, keyboard shortcuts for quick AI assistance, etc.

5. Awesome LLM Apps: Build awesome LLM apps with RAG, AI agents, MCP, and more to interact with data sources like GitHub, Gmail, PDFs, and YouTube videos, and automate complex work.

Hot Takes

1. Using Claude without MCP is like It's like buying an iPhone and only using it to make calls.

   Why would you limit yourself like that?

   It takes 10 mins to spin up and you'll outperform 99.999% of human beings on the planet. ~
   Greg Isenberg
   
   

2. Advertising will become invisible to people.

   Companies will start creating ads for your AI agent to consume, not you.

   We'll all have AI assistants that'll make decisions and purchase the items we need.

   We won't have to sit through a 30-second ad video anymore. Instead, companies will advertise in a language/manner that AI assistants understand. They will be the ones making purchase decisions. ~
   Santiago

That’s all for today! See you tomorrow with more such AI-filled content.

Don’t forget to share this newsletter on your social channels and tag Unwind AI to support us!

Unwind AI - X | LinkedIn | Threads

Awesome LLM Apps | Sponsor Us

PS: We curate this AI newsletter every day for FREE, your support is what keeps us going. If you find value in what you read, share it with at least one, two (or 20) of your friends 😉