Great to be back writing the blog. I have just recently (yesterday) updated the blog with a new design template. To my surprise, the switch was effortless thanks to Claude Code. Found a few templating stuffs I wanted to make changes for and Claude did this impeccably.

In actual fact, Claude Code has been a main stayer in my day-to-day coding work. I took out a 15 inch portable monitor just for it and positioned it below the primary monitor. The second monitor serves as the Claude Code terminal. This helped a lot, the focus now is bounded to the second monitor in comparison with multi tasking with the primary monitor only. The primary monitor now becomes a monitoring monitor of Claude Code changes.

I put a picture of the desktop setup as the cover image for this blog post. Please excuse the dust, it’s a given for when you live in the desert. There is also a David VS Goliath setup if you look closely: a Mac Mini M4 VS a full blown gaming PC. Actually the left side of the Mac Mini is where all my mini PCs are: an Intel NUC, an AMD APU NUC and a Raspberry Pi, all of them unused but I like the view.

A Couple Months of LLM Assistance

It’s good but not great yet.

Coding Agents like Claude Code is a big motivation, it replaced me almost entirely when done right. This made the difference that made Claude Code a main stayer:

https://github.com/SuperClaude-Org/SuperClaude_Framework

The /sc:* custom commands did what it promised to do. It has helped me debug, fixed bugs, write new features and build from scratch. I started regularly by challenging myself to only code using Claude Code for work. It was a whole new product that I’ve coded most of the foundational and core features. The next iterations I told myself must use Claude Code to have a taste of what it can or can’t do yet.

I ended up forking $100/month for a Claude Max subscription.

To be frank, I started the challenge with a more skeptical state of mind. But boy was I blown away. Here’s something I did for fun from scratch:

The following X post came from 2 people I have a lot of respect for. I follow this principle rather religiously, hence the second monitor.

I absolutely believe that LLM assisted activities is going to make our lives to the better kind of life we always want it to be. That is when we’re able to articulate our thoughts to the LLM and get the LLM to be able to articulate its thoughts with the kind of focus & clarity we want it to. The following is a snorkeling vacation with my son to Fujairah, UAE.

The 2 most used LLM apps for me are Claude Code and Claude Desktop. The tool usage with MCPs and the agentic operations of both are on a level of their own. The orchestration required to do this is an interesting technical question to answer. But more interesting for me is the product we could build with this kind of an experience.

This also happened though:

LLM models are going to get smarter and closer to AGI so the last tweet I think will be solved as time goes by. I also built an Android app for 1 audience, me:

I used the app whenever I go anywhere driving, it’s so cool to get context about an area I drove past everyday and also new areas I never been to.

The Audience Question

I liked the experience with Claude Code and Claude Desktop so much, I became an ambassador for them at work. The whole product and engineering team ended up getting a Claude Max subscription for each person.

Claude Desktop changed how products/feature are built at work. The product and business people uses Claude Desktop to ask questions to the codebase for the ideas they have in mind. Our founders used it to completely revamp how he sees operations can work, both of them reverted their decision to hire because now they see operations as a platform problem rather than a human problem.

That brings me to another point our founder pointed out; AI amplifies the output of humans. This aligns perfectly with everyone in the organization, we’re a small team and we try to keep it small across departments.

However there’s a glaring hole to be filled I think. All the above are only possible because I made the effort to setup each non-tech person’s Claude Desktop to be able to access our source code using the official Github MCP and a few other MCP tools. I think this is why Anthropic created Claude Code, they target tech people first so they become ambassadors to the company, after all Anthropic is a provider. This is effective, at work Anthropic models are first class in any AI efforts we do.

What if you’re not technical or you don’t know anyone who’s technical that can articulate how AI will help your day-to-day? There are plenty of AI startups out there optimized for narrow use cases for a specific domain but I don’t think that’s the answer. Nothing beats the chat UI’s generic capability to accept any query/question while at the same time being the UI with the highest accuracy to answer those queries/questions.

I want to solve this. Because I’m a software engineer, I am amplified by AI, I’m able to make my ideas come true with a velocity I never thought possible. It’s impactful and it’s here to stay.


I’m gonna wrap this up with a collection of tips & tricks for Claude Code. I’m off building stuffs.