2026/11/04

hmmmm…. days passed by and I gone through a lot.

Before that I was just using LLM, basically copy and pasting code. Things are still within control, I still handle technical details. Well… I think projects that I am doing are having low complexity, so I still can survive.

However, at the March 2026, I heard of my classmates discussing Antigravity, which is an agentic AI platform by Google. I got free subscription plan of Gemini from Google, so I plan to give it a try. I hereby built a hobby project (a timer app - Waypoint, meant to manage my time better). It is an important witness in my journey.

From March to November, I changed a lot (also my github). I switched tools from Antigravity to OpenCode, and now using Claude Code. WayPoint is turning more and more mature. The first version that AI generated is ugly, without any engineering discipline. It even got release keys committed directly to repo. Until now, at least it shows some technical rigor (reproducible dev env, modular code, well written documentation)…. ignoring the fact that I am the product manager (PM) and not actually doing any coding.

Often times I found myself struggling to do hands on -> used agentic AI and get mediocre work -> repeat. It saved me a lot. I can’t even imagine how devs survive during pre-agentic era….

This is me (which is committed to get rid of reliance to AI): img src

Well, well, well… Hence, I hereby declare that I am joining the cult. Agentic AI is the future. Thinking about my previous experience, I am that guy that performs better to learn concepts on paper than hands on. So, this era suits me, I even felt too lucky to go through these, to be born in this trend.

It felt destiny… Humans invented machines and now gyms are a replacement of manual labor, to the need of muscle training. Agents felt as same as machines… it frees developers from tackling small details, and we can now focus on the big picture. I guess the equivalent of gyms may be leetcode questions? At least it holds for me.

Its time for me to focus more on theories and use agents to apply those theories. I got a strong feeling that structure, understanding, and focus will be valued more in this world. I am both scared and excited for the upcoming days. Lets see how far I can go …

Written at Kampar, Perak - A place that is not my choice.

2026/04/04

Recently, I am pretty busy. Reasons are listed as follows:

  1. I have multiple subjects taken in this semester and didn’t really commit to understand the contents.
  2. I have tons of assignments, but not being motivated to even deal with any of them.
  3. I need to write a proposal for my FYP at next two semesters.
  4. I have been a collaborator in a project, where my laziness prohibited my capability to learn context in existing code, and contribute.

All of them actually reminds me of my incompetency, in this case which is about coding. It is AI enlarging my laziness (my fault) because it is so tempting to obtain answers for what you need to struggle for at least hours.

Hence, I was thinking of a structured method to remove AI from my current workflow. It is indeed feasible.

I wish to achieve this specifically for designing architecture and coding. I recalled that how I learnt Python (before OpenAI releases ChatGPT). That time, I blindly inserted many global variable just to let all my function execute. Indeed it was silly, but these silly mistakes build a strong foundation, which is indispensable for my growth.

Let me re-create this feeling and retain it, before placing myself in a production environment.

Hereby, I will limit myself to use AI to build subsequent open source projects (you can see the list in my resume), so that I achieve PBL by learning from AI instead of copying.

Further progress will be updated in ongoing this post.