All Dissections

Every article, deep-dive, and system design breakdown — organised, indexed, and searchable.

All AI & ML Algorithms Architecture Cost Reduction Design Systems Infrastructure Scaling Security

How Claude Code Connects to External Tools Through MCP

Claude Code is more than an AI coding assistant that reads files and generates code. Through the Model Context Protocol (MCP), it can connect to external tools, APIs, databases, and services, extending its capabilities beyond the local development environment. But how does this connection actually work? This article explores the engineering behind MCP, from its client-server architecture and tool discovery process to request execution and external API communication. It also examines how Claude Code handles authentication, permissions, security risks, and reliability when interacting with external systems, revealing how a coding assistant becomes a connected engineering agent.

Nov 6, 202620 min readRead →

Architecture

Inside DLD Toolkit For CSE KU: The Architecture Behind a Digital Logic Design System

DLD Toolkit For CSE KU is an offline digital logic design application built with C++17 and Qt 6. It brings Boolean analysis, circuit synthesis, and interactive simulation into one environment. This article explores the modular architecture, exact minimisation algorithms, event-driven simulation, and automated testing behind the application, revealing how engineering decisions shape its correctness, performance, and reliability.

Oct 9, 202613 min readRead →

Design Systems

Reducing AI API Fallback Latency: A Study Inspired by Cloudflare AI Gateway

Explore why fixed timeouts fail for AI APIs and how adaptive, speculative, and load-aware fallback strategies can sharply cut tail latency while controlling cost and overload risk. Drawing on Cloudflare AI Gateway patterns, this engineering study shows how to turn fallback from a simple emergency switch into a smart, measurement-driven control layer.

Oct 2, 202612 min readRead →

Cost Reduction

How ChatGPT Cuts Infrastructure Costs: The Engineering Behind ChatGPT's Battle Against Rising Compute Costs

OpenAI's ChatGPT triumphs over exploding compute costs through Mixture-of-Experts (MoE) architecture that activates only a fraction of its trillion-plus parameters per token, aggressive quantization, advanced KV cache management with paging and prefix caching, continuous batching for high GPU utilization, and speculative decoding. These optimizations, plus recent 2026 software wins cutting inference costs by over 50%, enable affordable, high-quality real-time responses at massive scale

Jul 24, 20268 min readRead →

AI & ML

Unveiling Cursor: How Advanced AI Agents and Custom Models Power Next-Gen Code Editing

Cursor represents a groundbreaking fusion of a Visual Studio Code fork and advanced AI agent technology, delivering an AI-native coding environment where custom models like Composer, semantic codebase indexing with vector embeddings, and speculative decoding enable lightning-fast, context-aware code generation and multi-file edits. Its agentic system leverages reinforcement learning, tool integration for file operations and testing, and parallel execution to autonomously plan, implement, and verify features from natural language prompts

Jul 24, 20268 min readRead →

Design Systems

Supabase System Design: The Platform That Powers Authentication, APIs, Storage, Realtime, and the Modern Backend

Supabase stands as a powerful open-source platform that redefines modern backend development by centering on PostgreSQL. It integrates essential services for authentication, APIs, storage, real-time updates, and edge computing while combining relational reliability with tools such as auto-generated APIs, JWT-based auth, global WebSockets, and serverless functions. This architecture ensures scalability, security, and portability for applications of all sizes

Jul 23, 20268 min readRead →

AI & ML

How Sora Generates Video: The AI Technology Behind OpenAI's Text-to-Video Model

OpenAI's Sora generates videos using a diffusion transformer (DiT) that processes spacetime patches of compressed latents, enabling coherent outputs with strong 3D consistency and object permanence from text prompts. At its core, the model compresses raw video into a lower-dimensional space via an autoencoder before iteratively denoising noise patches conditioned on detailed text embeddings to simulate realistic motion and world dynamics

Jul 23, 20267 min readRead →

Architecture

Why Netflix Moved from a Monolith to Microservices: A Comprehensive Analysis of Scalability, Resilience, and Innovation at Global Scale

Netflix moved from a monolithic architecture to microservices to fix scalability issues, single points of failure, and slow innovation as streaming demand grew. The original unified codebase faced a major 2008 outage, inefficient scaling, tight coupling, and slow deployments. By migrating to AWS and creating independent services, it enabled targeted scaling, fault isolation, and faster deployments.

Jul 23, 20268 min readRead →

Omegle’s Algorithm: Decoding the Technical Magic Behind the Internet’s Most Iconic Random Chat Platform

▶ VIDEO

Algorithms

Omegle’s Algorithm: Decoding the Technical Magic Behind the Internet’s Most Iconic Random Chat Platform

Omegle’s matching algorithm stood as a model of elegant simplicity combined with practical intelligence. At its core, the system maintained a global waiting queue—implemented as an efficient FIFO structure in memory or backed by high-performance stores—where new users were enqueued upon selecting text or video chat, with a background worker continuously monitoring length and instantly pairing by dequeuing entries and generating a transient session ID.

Jul 23, 20269 min readRead →

Scaling

How Bitly Scaled Technically: The Engineering Story Behind Billions of Redirects and a $100M+ ARR Platform

When you click a bit.ly link, it feels instantaneous. Behind the simplicity is a battle-tested system that scaled to billions of redirects through Base62 encoding, heavy caching, NSQ async pipelines, and a major MySQL-to-Bigtable migration. This engineering foundation powered Bitly’s growth from a viral URL shortener to a $100M+ ARR Connections Platform. The story showcases how thoughtful architecture turns simple ideas into resilient global infrastructure.

Jul 22, 202612 min readRead →

Security

How WhatsApp Encrypts Your Messages So Not Even WhatsApp Can Read Them

WhatsApp encrypts every message with Signal's X3DH handshake and Double Ratchet algorithm, so even WhatsApp's own servers never see plaintext. Here's how identity keys, pre-keys, per-message ratcheting, and Sender Keys work together to secure 1:1 chats, groups, and attachments end-to-end

Jul 22, 202620 min readRead →

Cost Reduction

How iLovePDF Reduces Infrastructure Costs: The Engineering Behind Its Low-Cost PDF Platform

Millions of PDFs are processed through iLovePDF every day, yet the platform remains fast, reliable, and surprisingly cost-efficient. This article explores the engineering decisions behind that efficiency, from hybrid cloud infrastructure and temporary file storage to load balancing, smart processing pipelines, CDNs, and feature gating. Learn how iLovePDF minimizes infrastructure costs while delivering a seamless experience to millions of users worldwide.

Jul 17, 202616 min readRead →

Showing 1–12 of 13 dissections