Kumar Anirudha's personal blog
This is the year everyone got a resident AI agent. Install one of the viral self-hosted assistants, give it a name, and it sits in your messaging apps around the clock, waiting for you to talk to it. The setup guides are refreshingly honest about what that costs: budget roughly 2GB of RAM per agent . One user running four of them on a 16GB VPS reported each agent idling around 1.5GB, spiking to…
If you have spent real time with a coding agent like Claude Code , opencode , or pi , you have probably noticed something. The model does the visible work, but the thing that makes the whole experience good is not really the model. It is the harness wrapped around it. The loop that keeps the model pointed at the goal, the context that shows up at the right moment, the tools wired in cleanly, the…
How a Rust physics engine compiled to WebAssembly quietly carries an entire walkable world, and how Runek builds those worlds on top of it. I’ve been building Runek , a source registry of procedural 3D components for React Three Fiber. Think “shadcn for 3D worlds”: you pull a component’s source into your project ( npx @runek/cli add bookshelf ) while the small runtime it imports ships as the…
I’ve a love hate relationship with rules. They’re the scaffolding I trust when everything else feels unsure. The reason physics works, the reason code compiles, the reason trains arrive ( sometimes ). Rules take the infinite and give it a shape we can actually hold. They let us say “if this, then that” and sleep a little easier at night. But rules are also the quiet fences we stop noticing. They…
Claude Code has a plugin system now. Which means anyone, including you, can bundle up a bit of AI-assisted magic, put a pretty bow on it, and have it installed by strangers on the internet in two commands. Writing the plugin is the easy part. Getting people to actually install it is the part nobody tells you about. This post walks through the whole round trip: build a Claude plugin, publish it to…
A Story You’ve Lived # It’s 4:47 PM. You open a pull request. It touches 12 files across 3 directories. CI is green. You tag one reviewer (the person who happens to be online) and close your laptop. Three days later, a bug makes it to production. Your reviewer missed a subtle interaction between two of those 12 files. Not their fault. They were one human reading a diff the size of a short story.…
Move fast and break things. That was the old motto. Now, things break even if you stand still. A founder I know spent six months building an AI writing product. Good team. Real traction. Paying users. Then one Tuesday morning, a foundation model company dropped a feature that did the same thing, for free, baked into a product with 100 million users. Six months of work. Obsolete by lunch. This…
Recently I went to an AI conference and saw dozens of people demo their agents. Agents that could search the web, summarize documents, write code, book meetings, deploy to staging, you name it. The demos were impressive. But every time someone in the audience pushed back with a slightly unexpected question or asked the agent to chain two tasks together, things fell apart. Hallucinated tools.…
If you’ve been building with LLMs for a while, you’ve probably built an agent at some point. You give the model a role. You add a few tools. You maybe throw in memory, some planning logic, and a carefully tuned system prompt. It works. Until it doesn’t. And when it doesn’t, it’s usually hard to explain why. At first, it’s tempting to blame the model. Or the prompt. Or the framework. But after…
You want to build an AI application. Something useful. Something your users can rely on. But here’s the problem: How do you know it’s actually good? You can build an app that generates summaries, answers questions, or writes emails. It works. Your tests pass. But is it reliable ? Does it hallucinate? Does it miss important information? Is it actually trustworthy? The philosophy is simple: You…
If you’ve been following the world of AI development, you might’ve heard the phrase “LLM-as-Judge.” It sounds dramatic, like some sci-fi overlord where one AI passes judgment on another. But it’s actually one of the most important evolutions in evaluating large language models (LLMs). LLM-as-Judge means using a language model itself to evaluate the quality of another model’s responses.…
A few months ago, as I was exploring machine learning while working on a project, one of my models kept behaving in a weird way. I had built a classifier to detect cats in images. During training, accuracy was awesome, near 99%. But in production, it failed miserably. The reason? Some of my “cat” photos had the same watermark on the corner. The model wasn’t learning cats, it was learning the…
Let me tell you about Vani. She’s a web development instructor teaching Node.js to a class of 30 students. Every semester, she spends the first two weeks just getting everyone’s environment set up. Some students have Windows, others have Mac, and a few brave souls are on Linux. Half of them struggle with npm permissions, a quarter get stuck with PATH issues, and there’s always that one student…
In this first article in this series, we went through all different rust plugins, the previous one , we understood the wasmrun plugin architecture. Here’s we’ll learn how to create your own wasm plugin which you can run using wasmrun . Creating wasm plugins, is not limited to wasmrun plugins alone, they can be quite useful in distributing as independent modules as well. What’s currently lacking…
In the last article , I covered all different types of rust plugin architecture and approaches we could have done. Now, lets dive into the plugin architecture used in wasmrun , an open-source WebAssembly runtime that supports multiple programming languages through a hybrid plugin system. Why Plugins in Wasmrun # Wasmrun intends to support multiple programming languages like Rust, Go, C/C++,…
Plugin systems are one of those architectural decisions that seem straightforward until you start implementing them. In the Rust ecosystem, plugin architectures present unique challenges compared to interpreted languages where you can dynamically load and execute code relatively easily. The Plugin System Challenge in Rust # Compiled languages like Rust require more careful consideration around:…
I’ve migrated the SOT (System Observation Tool) project from traditional pip/venv workflows to uv , Astral’s blazing-fast Python package manager. This is gonna be a game-changer for both contributors and users. Maybe I was just finding an excuse to use something from F1 movie fever. But speed is about performance as much as it’s about being quick. 10-100x faster dependency resolution and…
It’s the 2020s, and we’re all living in a world where the phrase “just ask the AI” has replaced “let me Google that”. From product design to poetry, from diagnosing diseases to picking your next Netflix binge, AI is no longer a tool we occasionally refer to. It’s the invisible co-pilot running the plane, the ship, and sometimes the planet. We are, in many ways, at the mercy of it. And mercy is a…
For the past few weeks since the first public release of Chakra , I’ve been really thinking about what the next right step for it is. I felt like driving myself into an overthinking well of right and wrong directions, I eneded up doing random stuff and going back and forth on updates. Also, not coding as consistently. The wasm ecosystem has really taken into the browser side of stuff while the…
The modern software landscape is shifting rapidly towards portability , performance , and sandboxed execution . At the heart of this transformation is WebAssembly (Wasm) . A low-level binary instruction format that allows code written in multiple languages to run safely and efficiently across environments. Initially built for the browser, Wasm is now powering: 🔌 Plugin systems 🧠 AI agents and…
If you’ve ever coded for the web, chances are you’ve touched JavaScript . But as apps get more complex, you might run into its quirks. That’s where TypeScript and AssemblyScript come into play. It all began in the wild early days of the internet, back in 1995 , when web pages were mostly just static documents. Plain text, maybe some images, and a lot of reloading. Netscape, one of the first web…
CPython is the reference implementation of Python written in C. When you run a .py file, it goes through several internal steps before your code is actually executed. Overview of the Compilation Process # Here’s a simplified overview of what happens when you run a Python script: Lexical Analysis (Tokenizer) Parsing (AST generation) Abstract Syntax Tree to Bytecode Compilation Execution by Python…
From browser booster to universal runtime, the journey of WebAssembly in 2025. A Bytecode Born for Browsers # In 2017, WebAssembly (WASM) arrived as a new low-level bytecode format. Safe, compact, and fast. It was meant to make games, 3D apps, and video editors perform better inside browsers. But developers had a bigger question: “What if this thing could run anywhere ?” WASI - Giving WASM a Real…
Python is one of the most widely used programming languages, but have you ever wondered how it works under the hood? CPython is the official implementation of Python written in C, and modifying it allows you to change Python’s behavior. CPython is the official and most widely used implementation of Python , written in C . When people talk about “Python,” they are usually referring to CPython. It…
Real-World Assets (RWAs) refer to tangible or intangible items from the traditional economy such as real estate, commodities, bonds, and intellectual property, that are represented digitally on blockchain platforms. Integrating RWAs into the Web3 ecosystem aims to democratize access to these assets, enhance liquidity, and introduce innovative financial products. Digital assets and real-world…
Imagine you’re deep into building something exciting. A groundbreaking app, a sleek dev tool, or even just a weekend side project that scratches a creative itch. You’re focused, in the zone, writing code and adding dependencies to move faster. After all, why reinvent the wheel when you can leverage existing libraries? But, here’s where things get tricky. Every package you install comes with a…
The way we represent ourselves online is undergoing a fundamental transformation. Traditional user profiles, tied to centralised platforms, restrict true ownership, portability, and privacy. But what if identities could be self-sovereign, interoperable, and decentralised , seamlessly working across all apps, chains, and ecosystems? We have to maintain separate profiles for each platform we use.…
In Part 1 , we built a LayerZero-powered bridge for moving NFTs between two EVM chains . But what if we want to bridge an NFT from Ethereum to Cosmos (e.g., Osmosis , Juno , or Stargaze )? For this, let’s go with the setup: LayerZero on the EVM side (Ethereum, Base, Polygon, etc.) IBC (Inter-Blockchain Communication) on the Cosmos side So, what are we doing? # EVM Side (Solidity) Locks the NFT in…
A cross-chain bridge is a protocol that allows assets, data, or smart contract states to be transferred between two or more decentralised networks. It enables interoperability between different blockchains, ensuring seamless movement of tokens, NFTs, and other digital assets across ecosystems. So, how about we learn how to build a cross-chain bridge that allows NFTs to move between two…
WebAssembly (WASM) is transforming how we run code in lightweight, secure, and cross-platform environments. Initially designed for browsers, WASM has now expanded into serverless computing and sandboxed environments . While languages like Rust and Go have robust WASM support, Python’s dynamic nature makes it challenging to run efficiently in WASM today. Why WebAssembly for Serverless Computing? #…
In today’s world, technology is often seen as a tool for efficiency, but in reality, it has become a structure of control. Operating systems dictate how we interact with software, app stores determine what we can access, and centralised networks govern our digital presence. The prevailing belief is that these structures are necessary for stability, security, and usability. However, don’t you think…
As the digital world transitions to Web3, decentralisation promises to redefine how we interact online securely, privately, and without intermediaries. At the heart of this revolution lies Decentralised Identity (DID) , a technology aimed at giving users complete control over their identity and interactions in the decentralised ecosystem. Let’s explore the world of DIDs and the impact it can have…
Liquidity pools in the context of traditional finance evolved over centuries, rooted in mechanisms designed to ensure that buyers and sellers can transact with minimal price volatility and at the right time. These mechanisms laid the foundation for the more modern, decentralized liquidity pools found in today’s digital finance (DeFi). In the early stages of financial markets, liquidity was…
In simple terms, bridges are tools that allow the transfer of assets, data, or even smart contract instructions between different blockchain networks. Think of them as the highways connecting isolated blockchain cities. For example, if you have some Ethereum (ETH) but want to use it on the Binance Smart Chain (BSC), you’d use a bridge to “wrap” your ETH into a token that works on BSC. The idea of…
Launching a crypto project comes with one big question: How do you sell your tokens? Token sales are the lifeblood of most blockchain projects, fueling development, building communities, and driving adoption. But picking the right sales model isn’t just about raising funds. It’s about fairness, accessibility, and strategy. From simple “buy it now” pricing to dynamic auctions and even gamified…
Imagine having access to the value of an asset, say, Tesla stock, without ever owning it. That’s essentially what synthetic asset tokens (or just “synthetics”) do. These are blockchain-based tokens that mimic the value of real-world assets, such as stocks, commodities, or even fiat currencies. Think of them as financial doppelgängers: they look, act, and feel like the original, but they exist…
If you’re looking to learn about the basics about Chain Abstraction, then read this article instead. Technically speaking, chain abstraction means abstracting away the differences between blockchain networks to provide a seamless experience. The goal is to: Simplify User Interaction: Make it easy for users to interact with apps without worrying about which chain they’re on. Enable…
If you’ve ever stumbled into the world of Asian graphic novels, you’ve probably come across the big three: Manga (Japanese graphic novels), Manhua (Chinese graphic novels), and Manhwa (Korean graphic novels). At first glance, they might all look like stylish black-and-white panels filled with expressive characters. But look closer, and you’ll find a rich tapestry of differences woven from their…
Say you want to make an online payment, and instead of just clicking a button, you have to figure out whether the store accepts Visa, Mastercard, or PayPal. Now multiply that headache by 100x, and you’ve got the current state of blockchains. This is where chain abstraction steps in — the fintech equivalent of a universal payment processor but for blockchains. It makes the Web3 world less about…
Let’s talk about chaos. Not the “forgot your wallet, keys, and phone” kind of chaos, but the deep, philosophical, scientific kind. The one that makes you question if the flutter of a butterfly’s wings in Bengaluru could really spark a thunderstorm in New York. Chaos theory is basically the science of surprise. Predicting the unpredictable . It’s like trying to organize your desktop files: no…
I was working on a side project, which led me to study about various network interfaces. Felt like some rabit hole I fell through. Before I knew it, I had listing down over 10 different network interfaces. So, thought of putting them together for somebody else also studying network interfaces. TCP/IP, or Transmission Control Protocol/Internet Protocol, is a framework that organizes the…
In the labyrinth of modern life, where individuality is often celebrated above all else, the idea of leading by example remains an esoteric yet vital concept. For the modern individual—a seeker of authenticity and a curator of their experiences—the power of example is not merely about imitation. It is about setting a precedent that inspires action, demonstrates clarity, and communicates intentions…
In the complex maze of the modern mind, where every turn presents a new fear, a new inadequacy, or a new ambition, lies the pervasive whisper of paranoia. Not the kind typically associated with conspiracy theories or spy thrillers, but a more subtle, insidious form that has become almost a default setting for human interaction in today’s society. Here, we explore the multifaceted nature of…
The web, in its earliest form, was envisioned as a decentralized platform—a space where anyone could share information without gatekeepers. Over time, however, the internet has become centralized. A handful of corporations now control vast amounts of online data, dictating how we interact, transact, and communicate. This centralization has led to issues like censorship, data privacy violations,…
In a world increasingly drawn to the pursuit of perfection, the notion of embracing flaws may seem counterintuitive. Yet, these so-called “flaws” might just be what make us remarkable. This paradox is beautifully encapsulated in the concept of The Perfect Flaw —the idea that imperfections are not just acceptable but also essential in defining authenticity, resilience, and growth. The concept has…
In an era where technological advances continue to break the boundaries of what once seemed like science fiction, we stand at a crossroads between the physical and the digital. The interface has always been the bridge between humans and machines—beginning with the simple command-line, evolving through graphical user interfaces (GUIs), and now, the immersive experiences of augmented reality (AR)…
A lot of the recent rise in isekai anime also suggests a favorable wind towards perfecting our system. In the anime Hunter x Hunter , the incentive system plays a crucial role in driving the narrative and character development, particularly in the context of the Hunter Exam and various arcs throughout the series. The Hunter Exam # The most prominent example of an incentive system in Hunter x…
As I started writing this, it became apparent as to how deep the rabbit hole really is… 🐰 So, I’m breaking up the discovery phase in multiple parts. If you’re just catching up, start at the beginning… 🏆 Incentive systems are frameworks designed to motivate individuals or groups to achieve specific goals by offering rewards, recognition, or penalties. These systems can be implemented in various…
Reentrancy attacks are one of the common security issues that can occur in Solidity smart contracts. They exploit the fact that a function can be called multiple times before it finishes execution . When a contract A calls a function in contract B, contract B can call back into contract A while it is still running. If contract A has not finished executing and is still holding state, contract B can…
Communication is a key in any ecosystem. Now with more cross-chain communication getting sorted using bridges, it’s high time to sort on-chain communications as well. Particularly, referencing of composable smart contract functionalities within the same chain. On-Chain Immutability # Blockchains in general are isolated networks. Smart contracts are even further isolated in the network given that…