Your cognitive load is the bottleneck This past couple of months, I’ve been working on projects. AI has drastically accelerated my learning and development. Questions to AI can start shallow (eg, “what is X?”) and go as deep as I want. AI more or less understands what I want to make, and urges me to write code step-by-step. I instruct AI not to give code and to let me write code myself first. And more often than not, the code I’m expected to write includes new items for me. ...

Today, I was discussing with DeepSeek why AI is not a super wise brain even though it has vast amount of knowledge available on the entire Internet. It’s answer mostly made sense; that it’s just a probability prediction machine and it doesn’t learn from experience. Looking at the attention mechanism which is the core of DeepSeek, ChatGPT, Gemini, etc., I understand what it means. It only cares about context/patterns. It forgets everything in a new chat without “memory” feature. It even forgets its training data unless repeatedly appeared, though I am often marveled at how AI can retrieve niche info from its knowledge. ...

I’ve been learning JavaScript and Vue (and re-learning Python) with Claude as my coding partner. I will share some findings and tricks when learning from AIs. Note: I’m (sort of) intermediate for Python and a novice for JS/Vue. AIs are trained to help, not to teach If you use general chat AIs such as ChatGPT, Gemini, Claude or Grok, they tend to give your full working source files from the beginning. This is not good for your learning. ...

I started using AIs heavily 2-3 months ago. For the record, I’m writing down my current AI usages. My current AI workflow Here is my AI workflow. I mostly use free-tiers. I feel $20 a month for an AI service is a bit too high although I know that they barely earn profits with the pricing. The only exception is Gemini API because I needed to top up $10 just to start using it. ...

It’s been 2 months since last post but this blog is not dead (yet). I’ve been busy working on two AI-related projects. Let me briefly introduce them. ragev https://github.com/achiwa912/ragev ragev is a RAG evaluation harness where you can change parameters such as top-k and evaluate performance. The README.org includes complete evaluation report. I added frontend UI to ragev because I wanted to use JavaScript + Vue that I had been trying to learn for a few years without success. Once again, I’m convinced that after reading an introductory book or two, I need to work on projects to be able to use a programming language. AIs are a great help at this learning phase as it catches all beginner mistakes I make and explain diligently to me without pointing out how novice I am. ...

This is a draft speech script for a Toastmasters club. Opening Do you use AIs? How do you like them? I used to dislike AIs. They are not good for the environment, they hallucinate and all that. But a few months ago, I was told at work that we all need to learn and use AIs to be more competitive and grow. So, I started learning AIs, of course, from AIs. What I did - followed I first learned how AI works, then advanced techniques to talk to AI, and I also did a PoC (Proof of Concept) AI project I asked AI if it can generate learning courses. The course schedules it produced were pretty impressive. So, I started them. The courses went smoothly, maybe a little too smoothly For example, in the PoC project, AI gave me everything; from steps to setup the environment, and to test scripts. I didn’t write a single line of code. And, they just worked! My expectations ballooned. I thought, “How far can I go with AI assisting me like this!?” But, I didn’t go very far. Because there were some issues. What went wrong - wasn’t learning 1st issue: AI gave me only shallow explanations. I needed to ask a lot of questions so that to fully understand and internalize the knowledge. 2nd issue: AI did literally everything - I was just its “hands” Steps to setup the test environment Working Python scripts AI even analyzed the test results I fell into the same trap as so-called “tutorial hell”. Tutorials are for complete beginners. Anyone can follow the steps and reach the goal. Tutorials give us basic understanding and a mild sense of achievement. But you never go deep with tutorials. After a few tutorials, you are still a beginner. After several weeks of training courses, I was still a beginner. “It felt like I was learning—but I wasn’t.” Why this happens - AI “behavior” My misunderstanding, or the “gap” came from AI behavior. AI just tries to help you in the chat session. It doesn’t push you to think AI doesn’t care about your wanting to learn and internalize something. It explains everything before you even ask. It gives you everything that you should do yourself. AI tries to make learning as smooth as possible, depriving you of opportunities to learn “AI isn’t optimized for your learning. It’s optimized for giving you a quick answer.” What I changed - Take control “initiative” So I changed one thing: I took control. First, I do the work myself. I just don’t accept answers. Second, I question AI, or more like interrogate AI. I ask more details. I ask for sources. And third, I verify, because AI can hallucinate. If you don’t take control, AI will just do all the heavy lifting for you. And you’re not learning. Conclusion - Reframe AI “tool” I’ve learned these the hard way, from failures. If you are an AI beginner, I hope you can avoid similar mistakes. Let’s enjoy learning new AI technologies. After thoughts Yesterday, I told ChatGPT that I need to deliver this speech today, and asked some advice. It told me that my script was too dense for a speech and thus I needed to re-write, which I stupidly did but I shouldn’t have. Today, I’m most unprepared. Thank you, ChatGPT. This speech was all about not to rely on AI, but I am doing the opposite.

To be honest, the word “Prompt Engineering” sounded like a bit black magic or even alchemy to me; It’s more of an art than a science. But as all IT engineers are required to improve our productivity using AIs, I decided to learn it and go a little deeper on the subject. This article is my learning notes as well as a practical guide for advanced prompting. In later sections, I have added insights that I gained through my prompt usage and Q&As with (or “interrogating”) AIs. ...

In the previous post, I learned how LLM works from LLM itself (Copilot, to be more specific). In this post, I will learn about RAG, of course, from Copilot. Methodology: Ask questions starting from “What is RAG?” to Copilot, and read answers Write down my understanding, asking Copilot if I’m correct Revise my writing until my understanding is solid Let’s start today’s learning. What is RAG? RAG (Retrieval-Augmented Generation) is a workflow for retrieving relevant external information and injecting it into user questions for LLM to consume. RAG supplements these LLM’s weak points: ...

I didn’t really like GenAI because it hallucinates, consumes lots of energy, has raised memory and SSD prices, etc. But as an IT engineer, I can’t ignore it. In this post, I’ll try to learn how GenAI (LLM) works by asking a lot of questions to AI. I mainly use Copilot because it is most lenient with hourly/daily usage limit. Below are mainly outputs from Copilot (but I modified/summarized them). Sorry if I didn’t remove all hallucinations. As it turned out, LLM as a technology is pretty interesting. Let’s see if I can learn something complex as LLM from AI. ...