Notes on software engineering, shipping products, and building fast, maintainable web experiences.
LLMs are today’s most visible NLP technology, but NLP is the broader field. LLMs sit inside NLP—alongside classic methods, evaluation, linguistics, and production system design.
LLMs look like magic because they speak fluently—but under the hood they’re probabilistic token predictors trained on massive text. Here’s what they are, what they’re good at, and where they break.
AI is changing the internet fast. Web3 can still be useful—but only in narrow, high-leverage places like provenance, identity, payments, and coordination.
AI is changing how people search and how results are delivered. Expect more zero-click answers, stronger emphasis on trust signals, and a bigger role for structured data and brand.
RAG is a pattern that lets an LLM answer with your data by retrieving relevant context first, then generating a response grounded in that context.
Model Context Protocol (MCP) is a standard way to connect AI assistants in your IDE to tools like files, Git, databases, and internal services—safely and consistently.