Free → Pro → Max in 3 Weeks: My Claude Journey
It Started With Curiosity
December 2024. A friend — Ketan Khairnar — told me about Claude. "Just try it," he said.
I'm a derivatives trader. I run a wealth management business. The closest I'd come to coding was writing Excel formulas. But I had a problem — every evening after markets closed, I was spending hours manually reconciling order books across three brokers. There had to be a better way.
So I signed up for Claude. Free tier. Just to see what the fuss was about.
Day 2: Hit My First Rate Limit
I asked Claude to help me write a Python script. I didn't know what Python was, not really. But Claude walked me through it — what to install, how to run it, what each line does.
By Day 2, I'd hit the free tier rate limit. Not because I was wasting messages. Because I was genuinely building something and couldn't stop.
Upgraded to Pro. $20/month.
Week 3: Hit Rate Limits Again
The Order Analyzer was getting complex. Multiple scripts, Excel formatting, weighted average calculations, trade cycle grouping. I was sending hundreds of messages a day, iterating, debugging, refining.
Pro wasn't enough. Upgraded to Max. $100/month.
Within 3 weeks of signing up, I was on the highest plan. Not because I wanted to spend more — because I couldn't stop building.
The Numbers
Since January 5, 2026:
- 23,747 messages sent to Claude
- 1,100+ conversations
- 163+ unique sessions
- Longest session: 32 hours straight — 1,108 messages in one sitting. That was the day I built the Client Performance Dashboard.
- Peak day: 4,300 messages across 6 parallel sessions
- Models used: Claude Opus 4.6, Opus 4.5, Sonnet 4.5
I didn't plan any of this. I just kept solving problems.
What Working With Claude Is Actually Like
People imagine you type "build me an app" and it appears. That's not how it works.
Here's what a typical session looks like:
- I describe the problem in plain language
- Claude writes code, explains what it does
- I run it. Something breaks.
- I paste the error back. Claude fixes it.
- I look at the output. "This isn't right — the P&L should be grouped by trade cycle, not by order."
- Claude restructures.
- "Still not good, give me something better."
- Claude delivers.
I've probably said "still not good, give me something better" about 200 times. And every single time, it delivered.
The key insight: you don't need to know how to code. You need to know what you want. Domain knowledge is your superpower. I know exactly what a trade cycle looks like, how P&L should be calculated, what a client report needs. Claude handles the syntax. I handle the logic.
The 32-Hour Session
The Client Performance Dashboard. A 13 MB self-contained HTML file with three tabs, interactive charts, embedded PDFs, glassmorphism design.
I started at night. By morning, the basic structure was working. By afternoon, I was tweaking chart animations. By evening, I was embedding invoice links. By the next morning, I was polishing the responsive layout.
1,108 messages. One sitting. My client thought I hired a design team.
I didn't plan to build for 32 hours. I just kept seeing what was possible and wanting to push further.
What Surprised Me
The iteration speed. In traditional development, you write code, compile, test, debug. With Claude, the feedback loop is seconds. Ask a question, get working code, test it, refine it.
Claude remembers context. Within a session, it knows every decision you've made. "Remember that Excel formatting we did for Sheet 1? Apply the same style to Sheet 2 but with green headers." It just works.
I started thinking in systems. After the first few scripts, I stopped thinking about individual tools and started thinking about workflows. The Order Analyzer feeds into the Dashboard Generator which feeds into the Client Dashboard. Claude helped me build not just tools, but a complete ecosystem.
The confidence curve. First week — I was asking Claude to explain every line. By week three — I was designing application architectures. "Use Tauri for the wrapper, React for the frontend, Python FastAPI for the backend." I don't write the code, but I understand what each piece does and why.
Tips for Non-Developers
If you're a trader, a business owner, a professional with zero coding knowledge — here's what I'd tell you:
Start with a real problem. Don't try to "learn coding." Try to solve something that wastes your time every day. The motivation to fix a real problem will carry you through the learning curve.
Be specific. "Build me an app" won't work. "Read this CSV file, group orders by trade cycle, calculate weighted average buy and sell prices, output to Excel with green for profits" — that works.
Iterate relentlessly. Your first output won't be perfect. Your twentieth version will be. Claude doesn't get tired of revisions.
Trust your domain knowledge. You know your business better than any developer. Claude knows syntax better than you. The combination is powerful.
Don't be afraid to say "I don't understand." Claude will explain it differently. Every time. Without judgement.
From Tool to Extended Brain
The shift happened somewhere around week 4. I stopped thinking of Claude as a coding tool and started treating it as a thinking partner.
"I have 813 days of trading data. What patterns should I look for?" Claude didn't just write analysis scripts — it suggested angles I hadn't considered. Averaging down patterns. Revenge trading after losses. The correlation between overconfidence and catastrophic loss days.
That project — the Trading Companion — changed how I see myself as a trader. And it was Claude who surfaced the insights.
What's Next
I've now built 10 projects. Desktop apps, web apps, dashboards, CLI tools, even a mobile app. Every week I find a new problem worth solving.
The barrier to entry for building software has never been lower. If you have a problem and the willingness to iterate, you can build anything.
I'm living proof.