The AI Tools I Actually Use Every Day

No hype, no affiliate links. Just the AI tools I reach for daily to run my businesses, write, code, and think.

ai technology building

Key Points

  • Claude Pro is my primary AI tool for writing, analysis, and thinking—it’s best at long-form work and complex instructions
  • I use a small stack (Claude, ChatGPT Plus, Claude Code, Cursor, Perplexity) that costs ~$100/month and saves me 8+ hours weekly
  • Start with one tool and master it before expanding—most people overload themselves with platforms they barely use

I’ve been using AI tools professionally for two years now, and I’ve watched the landscape shift from “interesting experiment” to “essential infrastructure.” But most people approach AI adoption wrong. They sign up for five new platforms, get lost in feature fatigue, and abandon them all. I did that too, until I dialed it back to what actually works.

Here’s my actual stack—no hype, no affiliate links, just tools I use every single day to run Rotate, write, code, and think through hard problems.

Claude Pro: The Anchor Tool

Claude is the foundation of everything I do. I have a Pro subscription ($20/month), and it’s the best money I spend on tools. Here’s why: Claude handles long-form work better than anything else. When I’m writing a detailed email to a business partner, drafting a newsletter for Ryan’s Roundup, or thinking through a complex technical problem, Claude is where I start.

What it’s great at: Following complex, multi-step instructions. Understanding context. Writing that actually sounds like me (not like generic AI). Code review and architectural thinking. Long documents—I’ve fed it entire business proposals and had it synthesize key decisions. Reasoning through ambiguous problems where the prompt itself is vague.

What it’s bad at: Real-time information. Browse the internet natively? Nope, you have to ask for it. Image generation—it can’t do that. And if you need something really fast (a few seconds matter), other tools edge it out. The latest Claude Opus 4.6 is powerful but slower than ChatGPT on simple tasks.

I use Claude for probably 70% of my AI work. If I had to pick one tool, it’d be this one. Hands down.

ChatGPT Plus: The Breadth Tool

I pay $20/month for ChatGPT Plus, and I reach for it when I need speed or breadth. It’s not my thinking partner—Claude is. But it’s my research partner.

What it’s great at: Quick questions. Image generation via DALL-E (I generate social graphics and website mockup placeholders). Web browsing—it actually works well when you need current information. General knowledge across a wide range of domains. Speed. If I ask ChatGPT Plus a straightforward question, I get an answer in seconds.

What it’s bad at: Complex, multi-stage reasoning. Editing my own writing—Claude’s just better at capturing voice. Long conversations where consistency matters. And the UI, frankly, is less conducive to the kind of deep work I do.

ChatGPT Plus is my second AI, not my first. I use it for research, quick fact-checking, and when I need to generate images. But for anything that requires real thinking, I’m back in Claude.

Claude Code: The Development Game-Changer

Claude Code is a separate product—you can use it if you have a Claude Pro subscription. It’s a code-native environment where you can scaffold projects, generate entire features, and have Claude see the full context of your codebase.

I use it for my side projects (Openmark, Refract, and various experiments). Here’s what changed: I can now build much faster. Ask Claude Code to generate a React component with Tailwind styling, and it builds it in seconds. Need it refactored? Done. Want a full Next.js project scaffolded? It’ll do that too.

The learning curve is real, but once you understand how to prompt it—what context to give, how specific to be—you stop thinking of it as “AI doing code” and start thinking of it as “a developer who’s extremely fast and never gets tired.”

What it’s great at: Scaffolding new projects. Generating boilerplate. Refactoring existing code. Building features when you know what you want but don’t know the exact syntax. Debugging when you can show it the error. Creating full-stack applications.

What it’s bad at: Understanding truly ambiguous requirements. It’ll generate code, but if you’re not clear on what you want, it’ll just invent something. It also struggles with very large codebases (anything over 100K tokens of code gets tricky). And you have to have a reasonable understanding of code yourself—you need to review what it generates.

I’ve built two projects entirely with Claude Code that I’ve launched. Not “mostly” built—built. That wouldn’t have happened without this tool.

Cursor: The AI-Native Code Editor

Cursor is an editor built from the ground up to work with AI. It’s based on VS Code, but it bakes in an AI assistant (you can use Claude or GPT models). I use it with Claude.

Here’s the workflow: I open a code file, highlight a function, and ask Cursor (with Claude as the brain) to refactor it. Or I describe a feature in a comment, and Cursor generates the code inline. It’s faster than bouncing between Claude Code and a separate editor.

What it’s great at: Quick edits. Inline assistance. Understanding your existing codebase while you edit. Debugging while you’re looking at code. The feedback loop is tighter than most other setups.

What it’s bad at: Large-scale project scaffolding. Cursor’s better for “I have a codebase and I want to improve it” than “I’m starting from scratch.” The interface can get cluttered if you’re not careful.

I use Cursor for my daily development work, especially when I’m working on existing projects. Claude Code is for the big picture; Cursor is for the day-to-day changes.

Perplexity: The Google Replacement

I stopped using Google Search about six months ago. Perplexity is better for most research questions because it gives you answers with sources, not just links. You ask a question, it searches the web, synthesizes the results, and shows you what it found.

I use the free version 80% of the time. The Pro version ($20/month) has been worth it occasionally, but I haven’t consistently paid for it.

What it’s great at: Research questions. Finding current information. Comparing options (best backpack for hiking, cheapest flights to NYC, which AI writing tool is most popular). You get answers with sources, not just a list of blue links.

What it’s bad at: Deep analysis. Complex reasoning. It’s faster than Google, but it’s not smarter than Claude. I use Perplexity when I need information, Claude when I need thinking.

For most people, Perplexity alone would replace Google. For me, it’s a complement to my AI stack—I use it for 15–20% of my research needs.

The Tools I’ve Tried and Dropped

I’ve tested dozens of AI tools. Here’s what didn’t stick and why:

Midjourney: Beautiful image generation. Stopped using it because DALL-E 3 via ChatGPT was faster for my use case. I’m generating social graphics and mockups, not fine art. If you’re doing professional design work, Midjourney is better. For me, it was overkill.

Notion AI: Seemed useful for notetaking. In practice, I didn’t need it. I use Notion less than I thought, and when I do, I’m usually just writing—not asking AI to help.

Jasper and Copy.ai: Specifically built for marketing copy. I tried both. The copy always felt generic. When I needed good copy, I went back to Claude. These tools are trying to solve a problem (faster copywriting) that Claude solves better anyway.

GitHub Copilot: People swear by it. For me, Cursor with Claude works better. Copilot’s real-time autocomplete is nice, but I’d rather have deeper reasoning and bigger refactors than faster inline suggestions.

Claude for Sheets and Gmail extensions: Built by Anthropic to work in Google Workspace. Downloaded, used once or twice, removed. The value-add wasn’t worth context-switching between applications. When I need Claude, I open Claude. I don’t want it embedded everywhere.

The pattern: I dropped tools that tried to be “just smart enough” without being truly useful. And I dropped tools that added friction instead of removing it.

The Math: Is It Worth It?

Here’s what I spend monthly:

  • Claude Pro: $20
  • ChatGPT Plus: $20
  • Cursor: $0 (free tier is solid, Pro is $10/month—I haven’t paid)
  • Perplexity: ~$3 (occasionally Pro)
  • Claude Code: included with Claude Pro

Total: ~$43/month.

I also spend maybe another $30/month on experiments—new tools I’m testing, subscriptions to things I use infrequently—but that’s not core to my workflow.

Is $50/month worth it? For me, absolutely. I’ve calculated that these tools save me 8–10 hours per week:

  • Writing and editing: 3 hours (Claude, ChatGPT)
  • Coding and development: 4 hours (Claude Code, Cursor)
  • Research: 1 hour (Perplexity, ChatGPT)
  • General thinking/problem-solving: 2 hours (Claude)

If you value your time at anything above $250/week ($50/week * 5), the tools are profitable. For a founder, consultant, or knowledge worker, that’s a no-brainer.

How They Connect: My Daily Workflow

Here’s how I actually use these tools—not in isolation, but as a system:

Morning (email, quick thinking): I open Inbox and ChatGPT for research questions, Claude for anything that needs writing (emails, feedback, synthesis). If I need context on something I’ve read, Perplexity.

Midday (writing projects): I draft in Claude. I ask it to refactor, tighten, and improve. For longer pieces like this blog post, I’m living in Claude—drafting, revising, asking it to maintain voice consistency. Maybe I pop into ChatGPT for a quick image or to fact-check something.

Afternoon (coding): I open Cursor if I’m editing an existing project. I use Claude Code if I’m scaffolding something new or need deep reasoning about architecture. If I’m stuck, I bounce between them—one for zoomed-out thinking, one for zoomed-in implementation.

Evening (side projects): Usually Claude Code for building something or Claude for thinking about what to build next.

The tools aren’t separate—they’re layers. Perplexity is the outermost layer (research). ChatGPT is the breadth layer (quick tasks, images). Claude is the depth layer (thinking, writing, complexity). Claude Code and Cursor are the build layer (actual creation).

This is important: I didn’t start with five tools. I started with Claude. Then I found gaps (image generation, research speed) and added tools to fill those gaps. I didn’t say “let me get one of everything.”

What I’ve Learned: The Real Mistake Most People Make

Everyone wants to optimize their tool stack. Understandable. But most people are optimizing the wrong thing.

The limiting factor isn’t which tool you use. It’s how well you use it. I’ve seen people with access to every AI tool under the sun who produce worse work than someone who knows Claude deeply.

Here’s my advice: Start with one tool. Get really good with it. Learn how to write prompts that work for you. Build a workflow around it. Then, only then, add a second tool if you’ve found a genuine gap. Most people never need a third.

If you’re just starting with AI, my recommendation is Claude Pro. It’s the most generally capable. It handles writing, coding, analysis, and thinking. Yes, there are tools that are “better” at specific tasks (ChatGPT for images, Perplexity for research), but Claude is the best place to start.

Then spend three months really using it. Read prompt guides. Build templates for recurring tasks. Ask it progressively harder questions. Let it be your thinking partner, not just an automation tool.

After that, if you find yourself repeatedly hitting the same wall with the same type of task, add a tool. But not before.

The Future

This stack will change. It always does. New tools will launch. Claude will get better. Cursor will evolve. But the principle won’t: use the tools that make you better, not the tools that feel like you’re working harder.

Right now, I’m genuinely excited about Claude Code. It’s opened up possibilities I didn’t have before—I can build faster, which means I can experiment more, which means I learn faster. That’s the leverage point I’m focusing on.

If you’re building a business or doing serious knowledge work, my suggestion is to carve out a few hours and try Claude Pro. Spend time with it. See if it clicks. If it does, you’ll figure out your next move. If it doesn’t, you’ve lost $20 and learned something about how you work.

That’s better than signing up for five tools and abandoning all of them in a month.

Want to go deeper? I’ve written about comparing Claude and ChatGPT, why AI won’t replace you but it’ll replace people who don’t use AI, and how to build a small business AI stack. I’ve also written a guide to prompting for non-technical founders if you’re new to this.

Tools mentioned:

← All posts