I am empowered

I used to feel a strange mix of anxiety and disgust every time I opened my code editor.

For months, using AI agents felt like a betrayal of the craft I loved.

I felt like I was losing the joy of programming, replaced by a machine that churned out logic I barely understood. It was a heavy, paralyzing shame. I felt like a fraud in my own codebase, staring at lines of code that I technically owned but didnt truly possess.

Finding my place in the loop

Everything changed when I shifted my perspective on who is actually doing the work.

I stumbled across a quote that hit me like a lightning bolt.

It suggested that we shouldnt say I wrote this (would be a lie to ourselves) or AI wrote this (it didn’t by itself tho), but rather

I wrote this using AI.

It sounds simple, but it changed my entire world.

I realized that while the AI can produce the raw text, I am the one accountable for the result. I am the guardian of the quality, the legibility, and the complexity.

The Ralph Wiggum breakthrough

My technical workflow finally clicked when I discovered new tools like Claude Code and the Ralph Wiggum methodology.

If you havent heard of it, its a way of running an AI agent that is much more thoughtful.

Instead of cramming every instruction and progress update into the models short term memory (context), it uses markdown files to track tasks. It treats your file storage like a long term memory bank.

This approach solves the massive problem of context bloat.

Usually, an agent fills up its memory with repetitive logs, which costs a fortune and makes the AI lose its focus.

By using these external files, I can restart the agent whenever a major task is done or when the memory gets too crowded.

Experts say keeping that window below 50 percent is the sweet spot for performance. With models like Opus 4.5 or Codex 5.1, that gives me about 100 000 tokens of pure, sharp focus to play with.

Ownership and the joy of the craft

Now, I feel like a leader again.

I can dive into unknown codebases and actually drive the project forward instead of just watching from the sidelines.

The secret is in the review process.

I make sure I understand every single line.

I check if I like the style.

If the AI is struggling with a specific logic puzzle, I step in and write it myself.

I run smaller edits now, which keeps me grounded in the reality of the software.

I feel like I can finally do stuff again.

The paralysis is gone because the shame is gone.

This is such an incredible time to be alive and working in this field.

Life has its tough moments and it isn’t always bright, but these moments of clarity and empowerment are worth cheering for. I am ready to rock my projects with a clear head and a mastered workflow.

See you in the next one, hopefully for something a bit less technical but just as heartfelt.