Personal blog by Samir Madhavan
Why a bet with positive expected returns can still wipe out an individual — the difference between ensemble averages (many people, one round) and time averages (one person, many rounds), illustrated with a Monte Carlo simulation.
I have a Raspberry Pi that's been sitting in a drawer for years. Pulled it out recently, popped in a fresh SD card with the latest Raspberry OS, and decided to actually do something useful with it.
AI isn't replacing product managers, it's raising the bar. The PMs who go AI-native will move faster, ship more, and become indispensable. The ones who don't will fall behind.
Tired of polite, characterless AI? Meet Gilfoyle AI—a custom agent with the nihilistic, deadpan personality of Silicon Valley's Bertram Gilfoyle.
Why most AI rollouts fail—FOMO-driven buys, FOBO-driven resistance, and option overload—and where the real opportunity lies for technical people who bridge the gap.
Using UN Comtrade data, AI agents, and a 3D globe to visualize trade surplus and deficit across countries—with some surprising insights.
A practical checkpoint on using AI and LLMs after two years of heavy use—what I use it for at work and in life, principles that actually improve results, mistakes to avoid, and where it’s delivered real impact versus hype.
A Duck Hunt-inspired game using webcam motion capture to squash bugs with your fingers.
A minimal, no-nonsense SQL client for data professionals who just want to write, format, and organize queries without the bloat. Built out of frustration with overcomplicated tools—samzerSQL keeps it simple.
When faced with 2000+ bot followers on X and no free solutions that work, I built a Playwright automation tool to detect and remove them using pattern matching—no paid APIs or broken extensions needed.
A simple manual workflow for extracting and organizing recipes from saved Instagram Reels. Since Instagram makes accessing bookmarks and copying text frustrating—especially on mobile—the process uses screenshots and an LLM like ChatGPT to parse and clean up the recipe text. The formatted recipe is then saved in a note-taking app (like Apple Notes) alongside the original Instagram link, making it…
A personal challenge to build and ship one open source project every week in 2026, leveraging AI tools and practical experience from implementing LLMs to create useful solutions.
Traditional compilers like C/C++ convert source code into optimized machine code for standalone executables, while Python compiles code into bytecode executed by the Python Virtual Machine (PVM). Python's flexibility comes from this portable, interpreted approach, with efficient module caching and optional performance boosts through JIT tools like PyPy.
Variable scope defines where a variable can be accessed within a program and how its name is resolved. Python employs lexical (static) scoping, where the structure of the code determines a variable's visibility, following the LEGB rule (Local, Enclosing, Global, Built-in)
Python manages variables and their scopes through namespaces and symbol tables. Namespaces organize names and objects across different levels like local, global, and built-in, while symbol tables, created during compilation, determine how and where these names are resolved. Together, they form the backbone of Python’s scoping and variable handling.
As a fan of Korean movies who hadn't explored Korean series before, I was intrigued by Raunaq's recommendation of "Moving," a 20-episode series that, after a slow start, captivated me with its mix of superpowers, politics, espionage, and emotional depth, making it a standout in a sea of content and truly immersing me in its fictional world.
If you are constructing a building, a blueprint of the building is mandatory before laying the foundation. It forms the basis for estimating the resources required, the number of construction workers, the time it will take to complete the construction and a direction that will guide the civil engineers. The same methodology also applies to digital products.
This analysis of Indian matrimonial data shows grooms (mostly engineers/MBAs) prefer younger, fair-skinned brides and seek "good looking, educated family persons," with notable gaps between self-reported body types (1% heavy) versus actual obesity rates (12-16%), revealing how India's $10+ billion matrimonial industry blends traditional values with modern credentials.
This analysis explores publicly available data from the TCS 10k run, visualizing patterns across age groups, gender, and finish times through interactive charts. Key insights reveal that the youngest runners have the fastest average times, older age groups show more serious dedication with improving averages, and while the fastest runners maintain an age advantage, the slowest finishers actually…
This data analysis project uses web scraping and social network analysis to explore Hollywood actor connections, revealing which actors have worked together most frequently and how any two actors are linked through their filmographies. Key findings include Christopher Lee having the most films among popular actors, Martin Sheen and Charlie Sheen being the most frequent co-stars, and Robert De Niro…