Your Turn:

Drop the vibe-coding techniques you have discovered so far that everyone should know.

Bookmark this thread – I'll update it monthly with new techniques as I discover them.

Why These Work:

Specificity beats generality: Structured prompts get better results

Interactive > One-shot: Debugging is a conversation, not a command

Context is everything: AI needs your project's specific patterns

Technique #6: Test-Driven AI Iteration 🔁

Instead of "fix this code," try:

"Write 3 failing tests that expose the bugs in this code. Then fix the code to make tests pass. Explain why each test was necessary."

Forces AI to understand the problem deeply before solving

Technique #5: Context Stacking 📚

Don't send isolated code snippets.

Template: "Here's the specific function: [code]. I need to [specific goal]."

AI gives much better solutions this way.

Technique #4: Context File Technique 📂

Create /.cursor/rules/context[dot]md in your project with:

Architecture decisions
API endpoints
Data flow patterns
Common patterns used

You don't need to reference it in every AI prompt. Code relevance automatically improves 4x

Technique #3: XML Context Prompting 🔗

Claude models respond much better to structured prompts using XML tags.


[your buggy code]



Expected behavior vs actual behavior



Framework: React, Database OR link .md file

Technique #2: YOLO Mode + TDD 🏄

Cursor's YOLO mode can automatically run tests and iterate on code until all tests pass.

Setup: Enable YOLO mode → Write prompt: "Write tests first, then code. Run tests and refine until all pass."

I've improved test coverage from 0% to 55% in a week using this.

Technique #1: The "Rubber Duck" Prompt 🦆

Instead of: "Fix this code"

Use: "I'm explaining this code to a junior developer. Here's what it does: [paste code or tag file]. What questions would they ask about potential bugs or improvements?"

Works 10x better for debugging.

The lesson?

Sometimes the best companies start by ignoring conventional wisdom.

Loved this story? Follow me for more threads on tech startups and product building.

Like & RT to share this founder journey 🚀

Growth numbers that shocked Silicon Valley:

• 2016: Public launch
• 2017: 20,000 users
• 2019: 1 million users
• 2022: 4 million users

From college dropout to building a $20B company in 10 years.

The breakthrough came from solving a simple problem: "Why do designers need to install software to collaborate?"

While Adobe and Sketch focused on desktop, Figma bet everything on browsers + real-time collaboration.

This changed everything.

"Browser-based design tools were the future."

Figma realized this after:
• 4 years of building
• 6 rejected investor pitches
• 3 complete product pivots

They kept going without giving up!

The first version of Figma was actually a FAILURE.

They spent 2 years building a drone design tool (yes, for actual drones)

But after burning through half their funding, they realized something crucial about product-market fit...

In 2012, Dylan Field was a 20-year-old college student.

He did something unusual: dropped out of Brown University after getting a $100K grant from Peter Thiel to NOT attend college.

Everyone thought he was crazy. But he had a vision...

Build errors are my nightmares