A Beginner’s Journey from Curiosity to Code
- sathyavenkatesh
- Oct 31
- 3 min read
Updated: Nov 10
As a founder, I’ve always believed that great products come alive when ideas and execution meet halfway. Yet, in most teams, that bridge feels wider than it should. Founders and product managers think in “why” and “what,” while engineers live in the “how.”That’s where my curiosity about Claude Code began . Claude is a conversational AI coding tool that lets non-engineers explore, experiment, and even prototype logic directly in plain language. For anyone building products, it’s like being handed the keys to the workshop without needing to be a mechanic. In this post, I’ll walk through how I got started, what I learned along the way, and why every founder and PM should try this at least once.
This is an ongoing series, so be sure to check back once in a while for next blogs!
Step 1: Pick the Right Model
Before diving in, you’ll need to decide which AI model works best for your workflow. For me, I wanted something conversational — a system that could handle follow-ups, remember context, and reason across multi-step instructions. Just as important, I needed the model to be comfortable with data ingestion — the ability to read and reason over structured inputs like CSVs, JSON files, or API data. That’s key for PMs and founders because most of our insights live in product dashboards, metrics exports, or customer logs.
I compared a few options:
Claude 3.5 (Anthropic) — strong contextual reasoning, great for multi-turn conversations.
ChatGPT GPT-4 — widely supported, robust plugin ecosystem.
Gemini Advanced — more visual, still improving with context handling.
For my use case, Claude’s conversational flow and file-handling made it the best fit.
Note: I am not recommending one model versus the other, I am just documenting what worked for my use case. Make your own judgement call on the basis of what you think works for you.
Step 2: Get Access
This is where it gets slightly tricky — you’ll need a Pro plan to unlock the better models. I decided to start with a one-month subscription on Claude.ai, mostly to test how it fits into my workflow. Can you get by without paying? Possibly, if you just want to tinker with smaller prompts or experiment on free tiers. But if your goal is to evolve from manual executor to orchestrator, then investing twenty dollars for a month of exploration feels like a small, high-ROI bet.
Step 3: Set Up Your Environment
You’ll need Node.js to run Claude Code locally.If you’re unfamiliar with Node, I’ve found a short primer — Understanding Node.js for Absolute Beginners — that explains what it is and why AI tools depend on it.
Once Node is set up, install Claude Code using this command:
curl -fsSL https://claude.ai/install.sh | bash
If you’ve worked with shell commands before, you’ll recognize this as a “fetch and run” pattern — it downloads the install script and executes it automatically. If you’re new to terminal commands, I recommend reading these resources:
A Quick Note on PATH
After installation, I hit a small snag:
“command not found: claude”
This usually means your $PATH (the list of directories your computer searches for commands) doesn’t include Claude’s install location. In simple terms, your computer can’t find where the app lives. The fix is simple — add the Claude install directory to your PATH environment variable. I’ll cover that and some common troubleshooting steps in the next post.
Coming Up Next...
In the next entry, I’ll walk through running your first Claude Code project, cloning from GitHub, and understanding how conversational coding can reshape how PMs and founders build products. If you’re following along, this is a good point to pause, grab a coffee, and make sure your install works. The fun part — actual coding — starts next.
Comments