~/writingls
Long-form write-ups dissecting system design and the trade-offs behind each implementation choice.
- backpropagating-ode.md
Backpropagating Through Time Itself: Reverse-Mode Autodiff of ODE Solutions
From what an ODE actually is, to how neural networks became continuous, to the adjoint sensitivity method that trains them without storing a single intermediate activation.
- rabbitmq-unpacked.md
How Messages Actually Flow Through RabbitMQ
RabbitMQ from first principles — why it exists, how it routes and persists messages, and what happens when things break.
- debugging-leetcode.md
How Platforms Detect AI-Generated Competitive Programming Solutions
A technical look at how contest platforms might separate human-written code from AI-assisted submissions — plagiarism detection, behavioral signals, cross-language similarity, and why false positives happen.
- rag-pipeline.md
Designing Production-Grade RAG Systems: Beyond Vector Search
Most tutorials stop at embeddings and a vector database. This is the architecture that comes after — continuous ingestion, hybrid retrieval, reranking, caching, observability, and the failure modes that only appear at scale.
- collaborative-editor.md
Lessons Building a Real-Time Collaborative Editor
From a 150-line Electron + Monaco prototype to a production CRDT-backed collaborative editor — the architecture, the failure modes, and where I'd start tomorrow.
- hello-world.md
Hello, World
First post — what this blog is, and why it exists.