👋 Introduction

Welcome to the inaugural issue of Declassified Technologies! I'm excited to share this journey with you as we explore the vast landscape of technology together. Whether you're a seasoned developer, curious coder, or tech enthusiast, this newsletter aims to introduce you to tools, technologies, and trends you might not have discovered yet, while keeping things grounded by also showcasing technologies that might be considered “legacy” but that are still widely used today.

This newsletter aims to be bi-weekly, avoiding overwhelming readers with frequent emails (there are plenty of daily AI or tech newsletters for those who have time to read that much), while still providing interesting information and highlighting topics the reader might not have known about otherwise.

Each issue will (try to) bring you:

• Deep dives into emerging technologies

• Fresh releases and updates worth your attention

• Hot topics from different ecosystems

• Recent trends that matter

• Interesting GitHub repositories

• Small updates across the entire tech ecosystem

This is the first issue I have ever written, so please bear with me 😃 
Feel free to share your thoughts and any feedback you may have by replying to this email.

Without further ado, let's dive in!

🔍 Deep Dive: MCP (Model Context Protocol)

Connecting AI Agents to other applications

You might have heard about the so-called AI Agents, which are supposed to be automated AI assistants that can do tasks on your behalf.

Model Context Protocol (MCP) allows these AI Agents to interact with other applications, like Github, Slack, Notion, a browser, you name it. It is an open source standard meant to standardize the way applications provide context to LLMs.

What's the Big Deal?

The standard was first introduced by Anthropic for its Claude LLM, but recently it is starting to be adopted by other AI giants like OpenAI, Google and others.

Before MCP, every integration between an AI system and an external tool required custom development work, creating what Anthropic described as an "N×M" data integration problem. If you had M different AI applications and N different tools or systems, you potentially needed M×N different integrations. This fragmented approach made it nearly impossible to scale truly connected AI systems.

MCP transforms this into an "M+N problem" by providing a single, standardized protocol. Tool creators build one MCP server for each system, while application developers build one MCP client for each AI application. The result is a dramatically simplified integration landscape that enables AI agents to seamlessly connect with databases, APIs, file systems, and virtually any external service.

There are already a lot of MCP servers on Github, and new ones are always being developed. Some of them include: Algolia, AWS, Azure, Grafana, Heroku, PostgreSQL, MongoDB, Puppeteer and others.

Core Architecture

MCP operates on a client-server model using JSON-RPC 2.0 for communication The architecture consists of three main components: Hosts (AI applications like Claude Desktop or Cursor), Clients (connectors within the host application that manage connections to MCP servers), and Servers (external programs that expose capabilities to AI models).

The protocol maintains stateful connections, allowing for persistent context and more sophisticated interactions compared to traditional stateless API calls. This stateful design enables AI agents to build upon previous interactions and maintain awareness of ongoing tasks.

How MCP Works in Practice

Claude Integration

Claude Desktop provides the most mature MCP implementation currently available. Users can connect MCP servers through native integrations (like Gmail and Google Drive), official MCP servers maintained by trusted developers, or community-built servers.

Setting up MCP in Claude is straightforward for basic use cases. For native integrations, users simply navigate to "Connected Apps" in Claude and enable desired services. For custom MCP servers, users can add them via the command line using claude mcp add <name> <command> syntax.

Once connected, Claude can seamlessly orchestrate complex workflows. For example, a user might ask Claude to "create a new release branch, run tests, deploy to staging, and send a notification," and Claude will coordinate actions across GitHub, Docker, Jenkins, and messaging platforms using their respective MCP servers.

GitHub Copilot and Visual Studio Code

GitHub has released its own MCP server as part of GitHub Copilot updates, marking a significant milestone in MCP adoption.

Visual Studio Code now has native MCP support in GitHub Copilot, enabling developers to ask Copilot to perform tasks that span beyond the editor. Developers can request actions like "Find any markdown files in my project missing an author footer, and create a GitHub issue to track adding those," and Copilot will autonomously handle the entire workflow.

Code Examples

Building an MCP server in TypeScript is remarkably straightforward using the official SDK:

import { Server } from '@modelcontextprotocol/sdk/server/index.js';
import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';

const server = new Server(
  {
    name: "notes-server",
    version: "1.0.0",
  },
  {
    capabilities: {
      tools: {},
      resources: {},
    },
  }
);

// Add a tool for creating notes
server.setRequestHandler(CallToolRequestSchema, async (request) => {
  const { name, arguments: args } = request.params;
  
  if (name === "create_note") {
    const { title, content } = args as { title: string; content: string };
    // Store note logic here
    return {
      content: [{ type: "text", text: `Note "${title}" created successfully` }]
    };
  }
  
  throw new Error(`Unknown tool: ${name}`);
});

// Start the server
const transport = new StdioServerTransport();
server.connect(transport);

Challenges and Limitations

Despite its promise, MCP faces several significant challenges. Security remains a primary concern, as MCP servers have broad access to systems and data. Community-developed servers may contain vulnerabilities or malicious code, requiring careful vetting before deployment.

Performance and scalability issues can arise with complex MCP deployments. The protocol's stateful nature, while beneficial for context preservation, can create resource management challenges in high-volume environments.

The current ecosystem is still fragmented, with inconsistent quality across community-developed servers. The planned MCP registry and compliance test suites should address these concerns, but their development timeline remains uncertain.

Bottom Line: MCP will certainly be a very useful standard that will keep growing in the future. It helps AI Agents evolve from isolated chatbots into capable agents that can meaningfully interact with the digital world.
If you have a tool that you use and want it to be integrated with your favorite LLM, check out the Model Context Protocol website.

🚀 New Release: Cursor 1.0

Finally a “stable” release of Cursor

Anysphere has officially launched Cursor 1.0, marking a significant milestone for the AI-powered code editor that's rapidly transforming how developers work. This major release introduces several groundbreaking features as it transforms the editor from a simple coding assistant into a comprehensive development platform with autonomous capabilities.

Key Features:

BugBot for Automated Code Review

BugBot automatically reviews pull requests and identifies potential issues before they reach production. When problems are detected, it leaves comments directly on GitHub with a "Fix in Cursor" link that opens the editor with a pre-filled prompt to address the issue. This seamless integration between code review and remediation significantly accelerates the debugging process.

Background Agent for Everyone

Previously available only in early access, the Background Agent feature is now available to all users. This cloud-based coding agent can clone repositories, work on separate branches, and push changes autonomously. Developers can assign tasks using natural language prompts and screenshots, then review results without leaving the IDE. The agent can handle multiple tasks concurrently, from UI fixes to content updates.

One-Click MCP Setup with OAuth Support

Cursor 1.0 simplifies the integration of Model Context Protocol servers with one-click installation and OAuth authentication support. This allows developers to easily connect Cursor to external tools like Notion, Figma, Sentry, Slack, and other services. The curated list of official MCP servers further streamlines the process of extending Cursor's capabilities.

Plus, the startup behind Cursor, Anysphere, has raised $900 million at a $9.9 billion valuation, making it one of the biggest fundings rounds in software history, so they are definitely doing something right. The company has grown to over $500 million in ARR and is now used by over half of the Fortune 500, including NVIDIA, Uber, and Adobe.

🔥 Hot Topics: Hyperliquid chain

Some blockchains are still thriving

Hyperliquid has emerged as a standout success story that's redefining what's possible in decentralized finance. This Layer-1 blockchain, built specifically for high-performance trading, has achieved remarkable growth and adoption metrics that rival traditional centralized platforms.

Stats

• It has over 481,000 users, even surpassing some crypto projects

• more than 50 billion trades while maintaining daily trading volumes that regularly reach $4 billion

• it commands nearly 60% of the decentralized derivatives market and has achieved over $9 billion in open interest

• monthly trading volume reached a record-breaking $248 billion in May 2025, a 51.5% increase from the previous month

This could be the infrastructure piece that finally makes blockchain apps feel as seamless as traditional web apps. Hyperliquid delivers a "KYC-free, CEX-like trading experience" that combines speed, intuitive user experience, and deep liquidity without identity barriers. If you are interested in derivative trading, check it out.

📈 Recent Trend: AI coding tools

Recently, AI coding tools are EVERYWHERE: AI agents evolved from experimental tools to production-ready systems.

There are now tools available for programmers and non-programmers alike that can help you build almost anything you want: mobile apps, games, entire web applications including frontend and backend integrated with real databases. Seems like the industry is just getting started, with new tools being released almost daily, some of which offering FREE plans.

Top Tools

1. Lovable.dev: An EU startup doing over $10M ARR (Annual Recurring Revenue) which has built 1.2M+ apps since launch (some of them might actually be good 😁 )

2. Factory.AI: Instead of AI Agents they use their own name called Droids, which they say “ship features 10x faster”. Time will only tell if this is true.

3. Rork.com: Vibe code a mobile app in minutes.

4. Windsurf: An agentic IDE, for advanced apps.

5. Heyboss.AI: Type one sentence and build a functional app in 9 minutes

The Shift

We're moving from "AI that helps" to "AI that does." Instead of suggesting code, agents now write all the code themselves starting from just a prompt. Wow! Who knows if the code these AI tools write will reach any production application, but it is certainly cool and helps prototype much faster.

Prediction: Probably by the end of 2025 most developers will work alongside AI agents daily, fundamentally changing how we build software.

I suggest checking out this incredible thread made by John Rush on X, showcasing 46! AI coding tools:

🏆️ Top GitHub Repo: Memvid

🌟3.1k stars+ and growing fast | Video-based AI memory library

Ok, this is a weird one. Memvid revolutionizes AI memory management by encoding text data into videos as an alternative to traditional vector databases.

How does it work?

🖼️ in short: QR codes

Built By Olow304 (Saleban Olow)

Quick Start:

# Create a new project directory
mkdir my-memvid-project && cd my-memvid-project
# Create virtual environment
python -m venv venv
# Activate it
# On macOS/Linux:
source venv/bin/activate
# On Windows:
venv\Scripts\activate
# Install memvid
pip install memvid
# For PDF support:
pip install PyPDF2

# Basic Usage in Python
from memvid import MemvidEncoder, MemvidChat

# Create video memory from text chunks
chunks = ["Important fact 1", "Important fact 2", "Historical event details"]
encoder = MemvidEncoder()
encoder.add_chunks(chunks)
encoder.build_video("memory.mp4", "memory_index.json")

# Chat with your memory
chat = MemvidChat("memory.mp4", "memory_index.json")
chat.start_session()
response = chat.chat("What do you know about historical events?")
print(response)

Make sure to check out the README file for more instructions if interested to deep dive into this.

🔄 Tech Updates

• xAI has partnered with Polymarket, The World's Largest Prediction Market. Grok can provide analyses based on X data

• Lingo.dev compiler is an npm library that makes React apps multilingual without rewriting the existing React components.

• We won’t get to AGI (Artificial General Intelligence) by just scaling LLMs. Plus Apple published a paper regarding Aritificial Intelligence’s Illusion of Thinking

• The company behing $USDC, the 2nd most used stablecoin (blockchain based USD), Circle, launched their IPO (Initial Public Offering)

🗝️ Legacy Revival

• Java turned 30 years old! And it is still used in one way or another in 90% of Fortune 500 companies

• The Symfony PHP Framework released version 7.3

• Laravel 12 was also recently released

• Some Wordpress extensions still support PHP 5.6 (end of life was in 2018), but not for long

• Python gets a new JIT (Just-In-Time) compiler

• Redis 8 has reached general availability

🐦‍⬛ X Hits

1. Fix your posture

2. Selling agents

3. Pump.fun (the memecoin trading platform) extrading every last drop of value from its users

4. New open source voice model

💡 Tech Tips for Next Issue

Got questions about any of these technologies? Reply and let me know what you'd like me to dive deeper into next time. I'm particularly excited to explore:

• Move programming language used in Sui, Aptos and other blockchains

• Real-world AI agent implementations

• Blockchain storage and security solutions

• Advancements in robotics and AI into the physical world

Thanks for reading! Forward this to a fellow developer who might enjoy staying updated on the broader tech landscape.

Till next time,
Rares.