I tested Gemini Pro 2.5 as my main coding model for 40+ hours.

Here're 2 documents that are working brilliantly well with Gemini.

"App flow document + App flowchart."

This made my Cursor workflow 10x better. Here's why it is working: ↓ Gemini Pro 2.5 has a 1 million token context window.

In simple terms it can scan upto 100,000 lines of code instantly to understand full codebase.

Claude & GPT models has 200k context window and that's why they struggle with large codebases.

Debugging is the most frustrating part of AI coding.

But it can be solved if you build a strong "wall of context" around AI models.

Here's how you can fix debugging in Cursor/Windsurf: ↓ If you say "build me a spotify like app" and you expect AI to code the entire frontend and backend in 1 prompt. That's not possible.

AI models can output limited code in 1 request.

Also, AI models are not perfect they have 2 limitations:
1. AI hallucinations
2. Context window

Google's Gemini Pro 2.5 is the best AI coding model right now.

I use Gemini 2.5 + Sonnet 3.5 + GPT o1 models in my new coding workflow.

After using it for 20 hours, here're my finding: ↓ I admit, Google has done a wonderful job this time with Gemini Pro 2.5 model.

They competed with Claude's Sonnet models for coding and the code quality is far better.

In new workflow I use:
- 80% of the time Gemini Pro 2.5
- 10% of the time Sonnet 3.5
- 10% of the time GPT o1

What is Vibe Coding?

Here's a complete Crash Course for you: 🧵 2024 has changed coding forever.

It gave birth to AI coding models like Claude Sonnet 3.5, GPT 4o that can code better than junior devs.

We saw AI coding tools like Cursor, Windsurf, Bolt, Replit, Lovable etc built on top of AI coding models.

But..

Truth: AI coding models are good for code execution, but not good at planning.

If you let AI plan next step while coding, it will mess up the codebase.

You can solve this by attaching a detailed "Implementation Plan"

Here's how I create my plan: ↓ When you build production-grade applications, you need to sure about:
- What features you want in v1?
- What features are out-of-scope right now?
- How's your UI will look like?
- What tech stack are you using?

These are basic but core questions that AI tries to figure out itself if you don't provide context.

I built 36 projects in last 14 months, with AI code.

Today, I'm sharing my Coding Docs with you.

I use these 7-docs in Cursor/Windsurf to build a knowledge base for AI models.

Bookmark this post so you can copy the templates. ↓

1. Project Requirements Document (PRD) These docs form a "knowledge base" for AI coding models so AI models car refer to them at any moment.

This Technique minimizes AI assumptions and hallucinations by 85-90%.

Here're other core docs that I attach:

2. App flow document
This docs is works like Google map for AI.

How to launch your first SaaS in next 21 days.

Complete Blueprint: ↓ 1. Rule #1: Build something that you can monetize.

If you have no ideas, here's how you can use AI to crawl the web and find you gaps.

Use:
- Perplexity Deep Research
- ChatGPT Deep Search
- Grok Deeper Search

Ask: "Find me 10 biggest gaps in XYZ industry that a new SaaS can fill."

"Where do XYZ industry spend most money? Give me a detailed breakdown."

"What are the main pain points of XYX industry. Give me a detailed list"

or if you have an idea, get some data on that:

"I have this idea.......... provide me a critical analysis on this idea. Is it worth building?

What are the top 10 competitors?

How big is the TAM (total addressable market)?

Evaluate my idea out of 10 and share your rational reasoning"

In 2024, I decided to build 12 startups in 12 months.

I ended up building 11.

→ 7 failed (never got sold)
→ 3 sold (on acquire .com)
→ 1 profitable (@CodeGuidedev )

In this letter, I'm going to explain how I sold 3 AI apps for $30k in total: I wanted to force myself to learn AI coding craft and build production-ready apps.

One day I came across acquire .com (a marketplace where people sell their SaaS apps)

What was surprising to me is, buyers were buying apps with 0 users and $0 in revenue, or profit.

When I researched deeper I found 90% of new startups are AI coded startups.

That was the motivational kick I was looking for.

So I decided....

This is "Glass Morphism" style.

I coded this interface in 10 minutes (one-shot), using:
- @CodeGuidedev (to generate coding docs)
- @v0 (to code the UI component)

Here's how you can design clean UI components: ↓ Design should be the most important factor of your app.

End user care about the experience. "First impression is the last impression"

Present you app in clean, modern and vibrant layout.

Step 1: Describe your idea and generate coding docs.

How to make Cursor Agent super powerful?

A complete guide: ↓ In future Cursor Agent will be able to run for hours and code the complete app autonomously.

We can see glimpses of this right now.

But there's a core issue, AI hallucinations!

AI models can make up things, hallucinate and start breaking the codebase.

We can't afford this.

After coding 36 app with AI models, I've compiled a security checklist that you can follow.

Trick: Paste this into AI coding tools (Cursor/Windsurf) and ask them to generate a security report.

Then fix all errors before launch.

BOOKMARK this: ↓ This happened recently!

Let's fix it for others.

I code with AI tools for 6-7 hours daily.

Built over 36 projects in last 12 months.

Truth: "Build me ........... app" in 1 prompt is not possible.

So, here're all the MISTAKES you might be making with AI code: 1. No Planning.

I go from idea to a well written draft for my MVP in few hours with my planning hack.

I just turn on ChatGPT voice and we have one-on-one conversation about what do I want.

15 minute chat then I ask: "write me a well structured draft on all the tings we've finalized in this conversation"

I use ChatGPT voice as my brainstorm buddy, critique, web researcher and then draft writer.

In the end I get the core features of this MVP on one page.

Don't build blindly. Plan before you hunt.

Do you want to use Cursor/windsurf for your existing Github projects?

Here's how you can make (existing or new) GitHub repo AI native.
↓ Cursor Agent/ Windsurf Cascade can overwrite code, delete files or duplicate existing files.

This can mess up the entire project and debugging on a large codebase can get messy.

So for existing codebases sitting on GitHub that's a huge risk.

Let's fix this core issue:

I am not a UI designer.

But with CodeGuide + Claude + v0 I've cracked the UI design workflow.

This design is purely coded with AI in 13 minutes.

Let me explain my NEW UI design workflow: ↓ First of all If you haven't read my last post on this topic,

Please read that so you can understand the basics: ↓

https://x.com/cj_zZZz/status/1899491125486116937

I built this UI design with v0 and Claude·ai (web app)

It took me 25 minutes to code this interface.

Let me share my new UI design workflow: ↓ Before we start here're few other pages of same web app:

- Sign up page
- Sign in page

You can see the style consistency.

Cursor/Windsurf technique to remove 80% hallucinations.

Stop AI to think and plan. Attach 'coding documents' and force AI to use it as a knowledge base.

If you code with AI, read this: ↓ When we let AI plan on it own, AI messes up the codebase.

Top AI coding models (Claude 3.5/3.7 Sonnet, GPt4o) are predictive models.

They can follow the the instructions and go from point A to point B.

But they are not good think, plan and map out next steps.

We can fix this.

I built this Interface entirely with AI in 47 minutes.

(without writing a single line of code)

I used 2 tools:
1. Claude AI (web app)
2. Cursor AI

Here's a step-by-step guide on how I built this app: 1. Write a detailed Project Documentation

If you want:
- consistent design
- high quality code
- less hallucinations

you need to write "project coding documents" to provide context to AI.

I use @CodeGuidedev to write docs, that includes:
- PRD
- app flow doc
- frontend guidelines
- backend structure
etc.

How I made my Cursor/Windsurf workflow ultra productive with "Context Boundary" Technique:

(Bookmark this for future reference) ↓ Even the Founder of @cursor_ai acknowledged that.

Let's dive into it:

New Cursor Coding Workflow: 🧵
(Bookmark this for future reference) 1. Unified Interface: Agent as Default Mode

The previous 'Chat' and 'Composer' tabs are replaced with a simpler dropdown menu:

1. 'Agent' mode is now the default mode
2. Use 'Edit' mode for precise changes
3. Use 'Ask' for planning and questions

It's much better UI/UX now.

Cursor/Windsurf has 3 major issues:

1. AI hallucinations
2. Loop of errors
3. Context awareness

To fix these core issues we need to build a strong {Context Boundary} around AI.

Here's how you can fix your AI coding workflow : 🧵 Cursor and Windsurf built around predictive models like Claude Sonnet 3.5 and GPT4o.

These models can only think about 1 possibility and that 1 possible step can be wrong 9/10 times.

We need to make sure that we don't let these models predict their next steps.

Cursor/Windsurf cheat code.

{Provide detailed context about your coding project}

It will minimize 85% of AI hallucinations and errors.

Here's how you can do it using ChatGPT: 1. Project Requirements Document.

Ask AI to write a PRD (for AI coding model) about your project.

It should include these sections.
- App overview
- User flow
- Tech stack & APIs
- Core features
- In-scope and out-of-scope

To make it air tight add these too:
- Non-functional requirements
- Constraints & assumptions
- Known issues & potential pitfalls

This doc will provide high level overview to Cursor/Windsurf