Blog & Essays

February 2026  ·  Computer Science · Compilers

Building the Bminor compiler was the most technically demanding project I've worked on. The course walked us from scratch — writing a lexer with Flex, a parser with Bison, a type-checker, and finally an x86 assembly code generator in C. Each stage exposed a different layer of how programming languages actually work.

The Hardest Part Wasn't the Code

I expected the assembly generation to be the toughest stage. In reality, the hardest part was keeping track of the invariants at every layer of the compiler. A small assumption made in the parser would quietly break things three stages later. Good abstractions and disciplined testing mattered more than raw cleverness.

What I'd Do Differently

Start with a thorough test suite from day one. Writing tests after the fact — especially for a compiler where bugs can hide in edge cases — is far more painful than building them incrementally alongside the implementation.

Overall, it's the kind of project I'd recommend to anyone who wants to truly understand how software works under the hood.

January 2026  ·  Finance · Algorithms

The IMC Prosperity challenge drops you into a mock financial market and asks you to write algorithmic trading bots in Python. You get to observe price histories, manage inventory, and compete against hundreds of other teams — all while the market's rules are only partially revealed.

The Market Teaches Humility

The first few rounds felt manageable — stable goods with predictable mean-reversion. Then basket goods, options, and goods with hidden influencing factors arrived. Strategies that worked one round would crater the next. The market punishes overconfidence quickly.

Statistics Over Intuition

The teams that did well weren't the ones with the most complex models — they were the ones who understood their edge clearly and sized positions accordingly. Trying to implement Black-Scholes on an options round was humbling; the fit was never as clean as the textbook suggests.

The biggest takeaway: uncertainty is a first-class input, not something to be eliminated before you start trading.