AI stack

Every tool has a job. No shiny extras.

This is not a list of AI tabs. It is the working setup behind how I research, think, write, and ship. Each tool has a role, so things stay intentional instead of scattered.

ChatGPT

Claude

Perplexity

NotebookLM

Copilot

Gemini

Gamma

OpenClaw

Why a stack instead of one tool?

Every AI tool is good at something and mediocre at a lot of other things. ChatGPT is fast but shallow on reflection. Claude is thoughtful but slower to iterate with. Perplexity finds sources but does not help you write. The moment you stop expecting one tool to do everything, your output gets better.

A stack is not about having more tools. It is about giving each tool the job it is actually good at, then connecting them into something that feels like a workflow instead of a guessing game.

The tools

What each one does and why it is here

ChatGPT

Execution

Daily driver for drafting, planning, and turning vague ideas into something usable.

Claude

Deep work

When I need deeper thinking, more nuance, or calmer long-form writing.

Perplexity

Research

Fast, source-backed answers when I need orientation or quick validation.

NotebookLM

Knowledge

Working with my own notes and documents so ideas stay connected to sources.

Copilot

Microsoft

The AI layer that fits naturally when I am already inside Microsoft 365.

Gemini

Google

Multimodal work and tasks that benefit from the Google ecosystem.

Gamma

Presentation

Converting ideas into decks, visual stories, and shareable structure.

OpenClaw

Personal AI

My always-on assistant, running locally and reachable via WhatsApp.

How they connect

A typical workflow, not a theory

Explore

Perplexity for orientation. NotebookLM if I have my own material to work through.

Think

Claude for depth and framing. ChatGPT for speed and iteration.

Build

Copilot inside Microsoft. Gemini for Google-native tasks. OpenClaw for local routing.

Ship

Gamma for presentations. ChatGPT for final polish and shareable output.

The build stack

What it takes to go from idea to live product

The tools above are how I think. The tools below are how I ship. When the Convergence Profile went from concept to working web app, it needed a completely different layer — not for research or writing, but for building, deploying, and running a product.

Claude Code

Development

The engine room. Command-line AI that writes, edits, and iterates on actual code. Where the Convergence Profile app got built.

Cowork

Planning

Where specs, brand frameworks, and build plans take shape. The thinking that happens before a single line of code gets written.

Vercel

Deployment

Push code, get a live URL. Hosting and deployment that makes going from localhost to production feel almost suspiciously easy.

Anthropic Console

API management

API keys, usage monitoring, cost tracking. When your product runs on AI conversations, this is the dashboard you check every morning.

Resend

Transactional email for the feedback form and newsletter. Simple setup, no enterprise email service headaches.

Google Cloud

Authentication

Account connections and identity setup. OAuth configuration and credential management for the authentication flow.

Design

Cowork and Claude Desktop for specs, frameworks, and stress-testing ideas.

Build

Claude Code for writing the app. Iterate fast, fix what breaks.

Connect

Resend for email, Google Cloud for auth. The infrastructure layer.

Ship + Monitor

Vercel to deploy. Anthropic Console to watch costs and usage.

Local setup

The part where AI stops being a browser tab

OpenClaw runs on my own machine. It connects WhatsApp to Claude, so I can message my AI setup from anywhere without opening a website. It sounds small, but the shift from "visiting a tool" to "messaging your own system" changes how you use AI entirely.

You → WhatsApp → OpenClaw → Claude → Reply

Want the full breakdown?

The AI workflow guide explains each tool, when I use it, and why. Practical, not theoretical.

Download the guide