My First Claude Code project
One day. One tool. $2K found.
Last Sunday, I made a spur-of-the-moment decision: I got Claude Code.
Jamie Ontiveros (thanks, bud) had raved so much about it that I had to try it.
I caved.
I knew the best way to learn was through practice, so I had to find myself a use case.
What could I do?
Then it hit me.
Earlier in the week, I ran into the perfect situation that could potentially be automated.
Allow me to explain…
As I previously shared in this post last year, I invested in some land in Brazil with my cousin.
Just a few days ago, we ran into a very unique scenario.
Some of the land we bought had overdue property taxes that, for various reasons, were being garnished from family members’ accounts.
Long story. Luckily, not mine, I should add!
Our purchase of the land was for all assets, including any liabilities. So if there was debt, we are on the hook for it.
Long story short, some of that garnished money had, again for various reasons, never been returned, even when the property taxes were paid off!
Weird.
But in weird, there are opportunities.
I thought…what if there are other scenarios like that, where funds were garnished, property taxes were later paid off, but the garnished money was never returned?
I thought I was onto something.
The problem?
This is an extremely manual process of clicking and typing through judicial websites.
But what if that could be automated?
Sounded like the perfect task for Claude Code.
At first, I just used regular Claude, explaining the idea, the websites I needed to visit, the order of each task, and so forth.
I then asked: could this be automated?
Sure enough, it could. Claude gave me code to try it out, and off I went, testing it in Python.
First time, it didn’t work.
I did some troubleshooting, fixed the code to adjust to the website’s frames, and… it didn’t work again.
More troubleshooting, more prompting, more questions.
Claude noticed there was a login required for a certain step, and… it failed again.
More troubleshooting, more coding, more prompting, and I found a solution around the login.
And… I failed again.
I troubleshot the issue probably 30+ times overall.
After nearly giving up, and being made fun of by my wife, I finally got it!
The kicker: Claude Code was the differentiator in getting better code.
267 cases reviewed in less than 15 minutes, showing about 18 of those cases as potential recovery opportunities. Net benefit? About $2K.
Hooked.
That’s one word I would use to describe my experience with Claude.
How about you?
Have you gotten yours yet?
Would love to hear your success (and why not failure) stories.