AI-First Building
Shaurya Bahl doesn't use AI as a supplement to his coding. AI is the primary building tool. Everything else wraps around it.
The Stack
| Tool | Role | |------|------| | Cursor | Main code editor. AI-assisted coding, inline generation, refactoring. This is where Shaurya spends most of his building time. | | Claude (Anthropic) | Thinking partner. Debugging, architecture decisions, writing complex logic, brainstorming solutions. | | v0 | UI prototyping. Generate front-end components fast, then refine in Cursor. |
This stack means Shaurya can go from idea to working prototype in hours, not weeks. The AI handles the grunt work -- boilerplate, syntax, repetitive patterns -- while Shaurya focuses on product decisions, UX, and business logic.
What Changed
Before AI
Shaurya spent three years learning Python. Then months learning JavaScript, React, Next.js through YouTube crash courses. Building a single feature could take days of Googling, reading documentation, debugging cryptic errors.
After AI
One prompt to Claude generates a working function. Cursor autocompletes entire components. v0 produces a full UI layout from a description. The same feature that took days now takes hours -- sometimes minutes.
"What took me years of classes, hours of coding, trial and error -- AI can now do almost instantly."
The 10x Factor
Shaurya estimates he builds roughly 10x faster with AI than he would with traditional coding. This isn't hyperbole -- it's the difference between:
- Writing every line vs. directing and refining generated code
- Googling every error vs. asking Claude to explain and fix
- Reading documentation for hours vs. having the documentation synthesized in seconds
- Debugging by trial and error vs. having Claude trace the logic and pinpoint the issue
For a solo founder who only has 2-3 hours after school to build, this speed multiplier is existential. Without AI, he simply couldn't ship at the pace he does.
How He Actually Uses It
Starting a New Feature
- Describe the feature to Claude -- what it should do, edge cases, how it connects to existing code.
- Claude generates a first pass.
- Paste into Cursor, let the editor's AI refine and integrate with the existing codebase.
- Test, adjust, ship.
Debugging
Instead of staring at error logs for an hour, Shaurya pastes the error into Claude with context. Claude explains what's wrong, suggests a fix, and often catches issues Shaurya would have missed.
Learning New Tech
When Shaurya needed to learn Swift for LockIn, he didn't take a course. He started building and used Claude to explain Swift concepts as he encountered them. The AI became a real-time tutor embedded in the building process.
Why It Works for Him
The key insight: Shaurya isn't a beginner blindly copying AI output. He has years of manual coding experience that let him evaluate what the AI generates, catch mistakes, and direct it effectively. The AI amplifies existing understanding -- it doesn't replace it.
A beginner using Claude gets code they can't evaluate. Shaurya using Claude gets code he can evaluate, refine, and ship. The difference is the foundation underneath.
The Democratization Angle
This is also why Shaurya runs AI + Frnds. He saw firsthand how AI collapsed the barrier to building. If a 15-year-old can ship multiple products with AI tools, anyone can -- they just need to know the tools exist and how to use them.
"The only thing that separates you from building your own app isn't skill, money, or experience -- it's just knowing how."
See also: AI and Coding | How I Write Code | Building Philosophy | AI + Frnds