Lei Feng 的 Blog

As tools like ChatGPT reshape how we work, prompting—the skill of designing effective instructions to guide AI behavior—has become a new critical skill. Techniques like zero-shot, few-shot, and chain-of-thought can improve outputs at a tactical level. But they cannot answer the bigger questions: which approach should you choose for which situation? and how can we collaborate with AI effectively? Read More...

In today’s digital world, terms like API and SDK are thrown around often. But what do they really mean? Here’s a simple breakdown using an analogy we all understand: a restaurant experience! 🍽️ Read More...

With the rapid influx of new LLM research papers, I wanted a way to store, summarize, and explore these findings seamlessly. The idea was to build a QnA system that could quickly retrieve relevant knowledge while keeping everything under my control. Given the excitement around DeepSeek R1, I thought this as a perfect opportunity to test its capabilities. In this blog, I’ll share how I set up a local Knowledge QnA system using Ollama, DeepSeek R1, and AnythingLLM—a solution that enables efficient querying of structured and unstructured knowledge sources while maintaining full data privacy and local execution. Read More...

In today’s information-explosion world, we are constantly bombarded with claims made by individuals, organizations, and companies. Advertisers emphasize the quality of their products, universities boast about the salaries of their graduates, and pharmaceutical companies promote the efficacy of their medicines. But how do we know if these claims are trustworthy? Hypothesis testing offers a systematic approach to assess the validity of such claims. Read More...

Have you ever got stumped by a logic puzzle? Sukodu got you scratching your head? Einstein’s puzzle made you crazy with the clues and facts? These puzzles might seem like mental gymnastics, but there’s a secret weapon can be used to crack the code: Constraint Programming (CP). Read More...