AI Building Tools Guide

A practical guide to AI tools for productivity and development. This is a living document—continually updated with new tools, use cases, and techniques.

Productivity Stack
Coding Stack
Tips, Tricks & How-Tos
When to Use What

Productivity Stack

AI Language Models (LLMs)

ChatGPT (OpenAI)

Best for: General-purpose AI assistance, writing, coding, and problem-solving
Ideal for: Daily productivity tasks, content creation, and brainstorming
Try it: chat.openai.com
Strength: Most widely used, excellent for conversational AI

Claude (Anthropic)

Best for: Complex reasoning, analysis, and nuanced conversations
Ideal for: Research, writing, and tasks requiring deep thinking
Try it: claude.ai
Strength: Superior reasoning and safety features

Gemini (Google)

Best for: Multimodal tasks, web search integration, and Google ecosystem
Ideal for: Research with real-time data and image analysis
Try it: gemini.google.com
Strength: Integrated with Google services and real-time information

Perplexity

Best for: Research with source citations and real-time web data
Ideal for: Fact-checking, research papers, and staying current
Try it: perplexity.ai
Strength: Always cites sources and provides up-to-date information

Research & Information Gathering

Multi-Model Research Strategy

Best for: Comprehensive research on complex topics using multiple AI models
Use when: You need thorough analysis from different AI perspectives
Approach: Use different models for different aspects of the same research

Grok Deep Research

Best for: Real-time information and trending topics on X/Twitter
Try it: x.ai
Unique advantage: Access to real-time social media data and trends

Note-Taking & Organization

Notion

Best for: Structured note-taking, documentation, and knowledge management
Try it: notion.so

Granola

Best for: Meeting notes and collaborative documentation
Try it: granola.so

Coding Stack

Quick Prototyping Tools

Best for getting ideas working fast, but not for complex systems

Replit

Best for: Fast prototyping, learning to code, and collaborative development
Ideal for: Developers who want instant setup with no configuration
Try it: replit.com
Limitations: Not suitable for highly complex systems or workflows

Base44

Best for: Rapid development with batteries included (pre-configured environments)
Ideal for: Quick projects that need to work immediately
Limitations: Less suited for ambitious, complex architectures

Bolt

Best for: Quick web app prototypes
Try it: bolt.new

V0 by Vercel

Best for: UI/UX prototyping and frontend component generation
Try it: v0.dev
Strength: Generates React components with Tailwind CSS

Advanced Development Tools

For ambitious, complex systems and workflows

Cursor

Best for: Professional development with AI assistance in your IDE
Ideal for: Complex codebases and ambitious projects
Try it: cursor.sh

Codex (Claude Code)

Best for: Agentic coding with Claude from the terminal
Ideal for: Developers who prefer command-line workflows
Try it: Check claude.com

Backend & Database

Convex

Best for: Real-time applications (chat, collaborative tools, live updates)
Ideal for: Projects that need reactive data and real-time sync
Try it: convex.dev
Strength: Excellent for real-time chat applications and collaborative features

Tips, Tricks & How-Tos

Cursor: Multi-Model Debugging Technique

The Problem: Stuck on a bug or complex implementation issue?

The Solution: Use multiple AI models in parallel

Open two Cursor windows side-by-side
- Window 1: Use Claude or ChatGPT
- Window 2: Use Gemini or Perplexity
Work the problem from both angles
- Present the same issue to both models
- Compare their approaches and solutions
- Use one model's solution to prompt the other for improvements
Pit them against each other
- When one model's solution fails, show the error to the other
- Have them critique each other's code
- Quickly iterate through bugs, package conflicts, and implementation issues

Why this works: Different models have different strengths, training data, and reasoning patterns. Using them in tandem helps you break through roadblocks faster than using a single model.

V0: AI-Powered Interaction Design Prototyping

How to Use AI to Prototype Interaction Design (Without Hiring a Designer)

This workflow perfectly illustrates why AI-assisted development is such a game-changer for polish and craftsmanship that wasn't feasible before.

The Process:

Start with one interaction you want to nail
- Pick a specific UI element or interaction you're trying to perfect
- Don't try to build the whole thing—just focus on getting one piece to feel right
Describe what you want it to do
- Tell the AI (like v0) what you're trying to achieve
- Example: "I want this element to fade out and zoom in at an angle."
Let it generate the first version
- The AI will create something. It might be close, might not be. That's fine—you're just getting started.
Now iterate like crazy
- Here's where it gets powerful. Ask for variations:
  - "Give me 20 different ways to do this interaction"
  - "I like the zoom on #3 but the fade on #8—give me 10 versions combining those"
- Keep going deeper on what feels right
- Within 3-4 minutes, you'll have 20 completely different approaches on a single page.
Pick the one that feels right
- Click through them, find the interaction that has the right emotional vibe
- Then drop that component into your actual code

Why This Changes Everything:

Before, you'd never ask a designer for 20 variants of a single interaction. The ROI wasn't there. The time cost was too high.

Now you can go infinitely deep on seemingly trivial details and add a level of polish and craftsmanship that just wasn't justified five years ago.

You're not thinking about cost. You're not coordinating with a designer. You're just able to experiment until something feels perfect.

That's the unlock.

You can prioritize feel and delight in ways that were previously reserved for products with massive budgets or teams.

This is what I mean when I say creativity will matter more than code. The barrier isn't technical anymore—it's about having the taste to know what feels right and the creative vision to explore until you find it.

When to Use What

Choose Your LLM Based on Task:

Use ChatGPT When:

✅ You need general-purpose AI assistance
✅ You're doing daily productivity tasks
✅ You want the most widely supported tool
✅ You need quick, conversational responses

Use Claude When:

✅ You need complex reasoning and analysis
✅ You're working on nuanced writing or research
✅ You prioritize safety and ethical AI
✅ You need help with detailed problem-solving

Use Gemini When:

✅ You need real-time web information
✅ You're working with images or multimodal content
✅ You're already in the Google ecosystem
✅ You need integrated search capabilities

Use Perplexity When:

✅ You need research with source citations
✅ You're fact-checking or writing academic content
✅ You want up-to-date information
✅ You need to verify claims with sources

Choose Quick Prototyping Tools (Replit, Base44, Bolt) When:

✅ You need to validate an idea quickly
✅ You're building a relatively simple application
✅ Time-to-first-working-version matters most
✅ You want minimal configuration and setup
❌ NOT for: Complex architectures, enterprise systems, or ambitious multi-component workflows

Choose Advanced Tools (Cursor, Codex/Claude Code) When:

✅ You're building a complex, production-grade system
✅ You need fine-grained control over your codebase
✅ You're working with existing, large codebases
✅ You want to be more ambitious with your architecture
✅ You need professional debugging and refactoring capabilities

Choose V0 When:

✅ You need frontend components or UI designs
✅ You want to prototype interactions quickly
✅ You're not a designer but need polished interfaces
✅ You're working with React and Tailwind

Choose Convex When:

✅ Your app needs real-time features (chat, notifications, live updates)
✅ You want reactive data that syncs automatically
✅ You're building collaborative tools
✅ You need a backend that "just works" with minimal configuration

Last updated: October 2025

Darren Buckner