Tron/Kits
// HOME / FIELD NOTES / THE JOURNEY INTO AI
← BACK TO BLOG

The Journey Into AI

2026-05-29 · TRONKITS FIELD NOTES

I have seen the future and the future is AI.

Since the beginning of the year, I have been diving head first into the AI pool.

I started off using ChatGPT and I was able to create some really cool utilities; but I was struggling with larger projects. I moved on to Claude chat, claude.ai. This was slightly better than ChatGPT, except Claude remembers absolutely nothing between chat sessions. Again, I was able to create some cool things. I was cooking with gas when I started using Claude Code. For larger projects Code is an incredible tool. Then I discovered Claude Cowork; Cowork is an awesome research tool, I used it to do the research for Climate50Index.com and RareEarthIndex.com. It did in a few hours, and a lot of tokens, what would have taken me weeks to research.

Claude Code isn't perfect; it has severe problems troubleshooting its own bugs. I spent 2 hours working with it trying to figure out why the database was getting 2 entries every time a submit button was pushed. Claude insisted that I was pressing the button twice, even when I pulled up the logs showing the entries were 10ms apart and no human can press a button twice in that period of time. All the troubleshooting was centered around preventing the button from being pressed more than once. I could not break it out of that track. Finally (I don't know why I waited that long) I opened up the browser debugger and stepped through a button click, the problem was very obvious, there were 2 listeners for the button click. During a re-write of some code, Claude did not delete all the "left behind" code. I asked Claude "should there be 2 listeners"? It came back, "no, that is the problem". It went on taking credit for finding the bug, LOL. This same exact scenario happened 2 more times, I was pretty quick to tell Claude to check for extra listeners.

You really have to work your way around memory issues with Claude. You must be methodical about including things in an "md" file that Claude will need to know. Things that you forget will bite you, like building a complete application in the wrong framework. Yes, I've done that.

It takes a while to get used to the way Claude distributes tokens. You get so many tokens every 5 hours, and so many tokens a week. I exceed the 5-hour limit pretty regularly, and have bumped up against the weekly limit a couple of times, and I'm on the "max" plan. It's not a deal breaker — you just learn to plan around it.

Here are a few of the projects I built since the first of the year:

I have published over 50 web sites in the past 5 months. Many of these are simple one page web sites, but several are full blown applications. I wouldn't have been able to create a quarter of these without AI assistance. When you get used to the tools you can become incredibly efficient. While the LLM is thinking through a complex task, you can be working on something else. I'm moving between screens all the time.

I hear a lot about companies laying people off because of AI. This makes no sense to me. If the people you have are making money for you, and now they can make four times more money for you, why would you lay them off? The thinking is all backwards. AI shouldn't be replacing people, it should be making them better, so they can make you more money. You have to think of AI as a tool, it's not magic, it's not the boogie man, it's just a tool.

OK, I'm getting off my soapbox now. I have worked with Gemini, Claude and ChatGPT, they all have amazing capabilities. Imagining life without AI is like imagining life without a cell phone, I could do it, but why would I want to.

The future is here.

I, George Clay, wrote this post in whole. AI was used to clean up my grammar/spelling, and point out any gaps in continuity.