# A year of coding by talking: what I gained and what I lost > Source: > Published: 2026-08-26 06:35:29+00:00 2024 was the year AI was everywhere. The ads, the praise, the feeling that something had already been decided without me. I am in my fifties. I had to decide whether to watch or take part. I decided to take part. I started with VS Code, on a paid plan. I did not have to think about what to build. Something had been sitting in my head for years: an automated trading system. I have lived fifty years, and while raising children the money got tighter, not looser. Financial freedom was moving away from me, not toward me. So I wanted to make money with automated trading. I think the idea first arrived in my mid-forties. That is why the decision to take part came so quickly. **Whatever I said out loud would simply get built.** That was the hope I walked in on. I do not really know how to code. But the AI would handle that part, so I trusted it. A year inside VS Code taught me that two names mattered: GPT and Claude. I used them in turn. I used them one at a time. Two problems. First, even on a paid plan the usage ran out fast. Faster than I expected. The road ahead was long and I was sitting still, waiting for a quota to reset. The second one was worse. **The explanations were excellent. The results were not.** That is where the illusion cracked for the first time. I still could not let go, so I paid for more. Adding Cursor bought me some headroom. And a different problem showed up immediately. **Change the model, and it wants to rewrite everything from the beginning.** Handed code written by a different AI, it would rather replace the whole thing than edit it. That is when I understood that switching AI mid-project is a bad idea. Everyone talks about pricing. Almost nobody talks about this one. And this is the one that actually held me back. I spent a lot of time fighting the tool. In the end I paid for Claude's hundred-dollar plan, and from then on I worked with Claude. I worked on trading at home, whenever I had time. Then at work I ran into a repetitive data-entry task. Other people usually handled it, and all of them found it painful. When something got entered wrong, the stress was real. I took my turn at it too, and I hated it. That is when Claude came to mind. Not really believing it would work, I said I would try to automate it, and started building. This did not go smoothly either. As the code got longer, it kept showing me that **it no longer knew what it had done.** So I decided to split the files, and asked Claude to split them. That finished quickly. That is where the trouble started. **It started working without knowing what was in which file.** The same change would land in several files at once. It sprawled. I told it to check the files. Every new session it fell apart again. It kept happening, so **I merged everything back into one file.** I did that out of pure frustration. By then the code was five thousand lines. Writing documentation did not occur to me at that point. Merging was the only move I had. The automation worked in the end. What was left behind was a single file of more than ten thousand lines. I would not recommend it to anyone. It was simply the best I could do at the time. That was where I felt the limits again, and started looking for another way. **Maybe the AI is the one that needs the standards.** That thought started here. The whole thing took about two months. It felt like a small success inside a beginner's failure. I still do not have it. Every time I thought it was close, something came back to tell me it was not. The bigger problem was somewhere else. **There were many tests. There was a great deal of verification. All of it stood alone. None of it was ever brought together.** Claude would tell me it was all there. What was there were test results. **There was no integrated file that actually traded.** That one was heavy. I still could not let go, because the dream of it keeps circling in my head. I only saw it while writing this. What I built and what I failed to build had **the same cause.** | outcome | what actually happened | | |---|---|---| | the data-entry automation | worked | I put the scattered pieces back together | | the trading system | did not | I never managed to put them together | Same illness, twice. One of them I could reverse. The other I could not. So this is where I have landed. **The AI needs standards, and it needs records.** Not because it is bad at its job, but because its limits are clear enough that I should be the one preparing for them. That is a statement about what I have to do, not a complaint about the tool. This series is the record of that year. It is not a success story. It is an account of what the AI was looking at, what I was looking at, and why the two kept failing to line up. Will this ever work? I still do not know.