What this covers: Why I switched from ChatGPT to Claude as my primary AI, what makes them different, and how to move your context over if you want to try it

Read this if: You're curious about AI alternatives beyond ChatGPT, you want AI for thinking (not just executing), or you're wondering what else is out there

You'll walk away with: A clearer picture of the AI landscape, criteria for choosing tools that fit your work, and the exact process I used to switch without losing my ChatGPT history

Jump to:

Professionally I'm obliged to give the diplomatic answer: "As an AI practitioner, I'm model agnostic. I use all of the AI models (yes there is more than ChatGPT) to further my practice and support my work. Each has its own strengths and advantages."

And it's true.

But, just between you and me: I really fucking love Claude.

A lot of folks are surprised by this. Myself included. When I first encountered it two years ago, I thought it was fine. But not fine enough to get me to use it primarily over ChatGPT.

And some of you inevitably ask: Well, why would I even bother using another AI? ChatGPT knows me so well.

So I'll put it this way:

If you're tired of ChatGPT's voice, the way it kisses your ass, or are a little suss on the direction OpenAI is taking, please know there are other options.

That's especially the case if you want to use AI to help you think, not just execute.

There's Gemini from Google (a separate piece on this is coming). There's Co-Pilot. There's Mistral. There's DeepSeek. There's Grok. And then there's Claude.

Even I'm overwhelmed by the AI hype and news. So, I'm not interested in contributing to the noise. This newsletter is about sharing my experiences in how to think well with AI and analog methods.

My business requires me to use a variety of AI. But, Claude is the tool I use the most both professionally and personally not only because it fits my needs, but because I genuinely like using it.

Everything here is experience and opinion.

Claude is like ChatGPT. It's an AI chatbot that you message and it talks back to you. ChatGPT is made by OpenAI, while Claude is made by another company called Anthropic.

Think of the different AIs like different car makes and models. Toyota makes 4Runners, Corollas, and Camrys. Each has different packages: the base package, the sports performance package, the premium package. Different cars serve different needs—some prioritize fuel efficiency, others off-road capability, others luxury. You choose based on what matters to you.

OpenAI has ChatGPT that gives you the option of auto-thinking, deep-thinking, in 5.1 and 5.2 (as of publish date) etc., and Anthropic has Claude with its different models.

To be clear, you cannot access Claude via ChatGPT or the other way around. If you have a Toyota Camry, it doesn't mean that you have access to a BMW X5.

Anthropic is much smaller than OpenAI. ChatGPT has 800 million monthly active users (free and paid), while Claude has 19 million (that's 1 Claude user for every 40 ChatGPT users). That's probably why most folks haven't heard of Claude.

You can use Claude primarily through a variety of products:

First, Claude.ai, AI chatbot via the web, desktop, or mobile app. There's a free tier and paid options ranging from $20-$200, which is on par with OpenAI ChatGPT.

Within Claude you can access three models (as of January 10, 2025): Haiku 4.5, Sonnet 4.5, and Opus 4.5.

There are also APIs more technical users can use to access these models. APIs are for automations, workflows, building apps, etc. For this audience and newsletter, I'll say that Claude.ai is powerful enough to use.

If you use Chrome, the Claude extension is another great tool.

It's not its own "agentic browser" the way ChatGPT Atlas or Microsoft Co-Pilot on Edge or Perplexity's Comet is. (Agentic browsers are basically AI-assisted web browsers that allow you to delegate tasks and actions to an AI that can read your tab. It can add things to your shopping cart, synopsize content on a page, create dashboards from information on a page because the AI agent can see what you see—and then some).

And there's Claude Code.

Despite the name, it's actually quite helpful for non-coders. Many folks have asked me about it. There's a lot for me to say so I'm working on a series of my non-technical self learning to work with it. For now, I'll leave you with this: there's a learning curve.

Unless you understand what I mean when I say "you have to access it via a CLI such as your terminal" you may want to slow your roll before going from zero AI experience to Claude Code. Learn from my pain.

A better start is a paid Claude account because:

[EDITED January 14, 2025] But wait, there’s also Claude Cowork!

Right after this article officially published, I got beta access to a new feature called “Cowork.” It’s basically a hybrid bringing together the chat interface of the Claude AI with the power of Claude Code. TL;DR it’s a user-friendly way to get a taste of Claude Code. You now have an AI agent that can do stuff for you, not just tell you about it.

What it does: It focuses Claude on the “doing” part without the intimidation and (potential risk) of working in a command line interface. You chat with it, it has access to tools and other apps online and files/folders on your computer. (Anytime you hear/read “it’s run locally” means it’s run on your own computer). Analyze a bunch of data in a folder. Re-organize your desktop. Generate a spreadsheet or deck with data from your computer. And then some.

Folks are still experimenting - myself included. It’s still very rough around the edges and not completely integrated with Claude’s full UX. But considering that the Anthropic team essentially built this in 1.5 weeks, I’m impressed. Great example of trailing intent - noticing how users are using the product and evolving it from there.

But here’s why I’m excited about this as thinker: It’s a glimpse of what’s to come.

Knowledge workers have a more user-friendly interface to do, not just chat. And while Microsoft, Google, and ChatGPT have focused on agentic-browers. Now, we’re seeing the power of agentic support for files locally.

How does this affect thinkers? You’ll get to do/try/experiment more. And it also signals you’ll have to think better upstream (a trend I’m seeing everywhere) to get the best results from these capabilities.

1. You want to use AI as a thought partner (and maybe you’re a control freak).

I mentioned earlier that Claude just hits different. Anthropic has designed its models in a way that it checks in with us more.

What does that mean? In ChatGPT, you may have noticed that especially with the rollout of ChatGPT 5 and 5.2, it's really eager to do things for you. It's agentic, it wants to execute tasks.

In my case, I wanted to debrief on a difficult professional conversation I needed to have. Using the same prompt and conversation starter, I went to both ChatGPT 5.1 Thinking (new at the time) and Claude Opus 4.5 (thinking model equivalent).

ChatGPT came back with words of encouragement, some insights based on my context because it has memory of me, and also spun up an action plan on "how to have difficult conversations."

Claude on the other hand, engaged me in a 20-minute back and forth, probing for context and asking better questions. By the end of it, I felt more prepared with receipts and riz to engage professionally with this stakeholder.¹

My takeaway: With some course-correcting, ChatGPT can be a great thought partner. But Claude is a more powerful thought partner that can also be a great assistant in delegating tasks like generating a slide deck, creating spreadsheets, Google Docs, etc.

That's why I tend to recommend Claude to all of my clients who are knowledge workers or founders/entrepreneurs who may not have peers or colleagues at the ready to discuss and work through problems that they uniquely have.

All have reported back how helpful Claude has been in working around messy first approaches to situations, new topics, complex problems and convos requiring thoughtfulness not just on the AI's part, but their part—the human.

2. You write a lot and maybe need a reality check, too.

AIs are designed to be helpful. In the case of ChatGPT, like really helpful.

Claude can certainly be blindly leading your hype squad, too. But I find it less so than ChatGPT.

The difference I've noticed is that Claude is good at identifying patterns AND actually calling you out on them, going so far to even stop and pause the convo.

For example: I use Claude often to talk through business cases for my consulting as well as strategy around the media arm of my business. And as I tend to, I was rabbit-holing/spiraling on one issue.

That's when Claude came back with: "Before I answer this, can I ask you, are you discussing this because you believe it has impact, or are you discussing this because of some sort of fear of [redacted]?"

Claude called me out.

One thing people mentioned about Claude early on is how it doesn't have "personality" the way ChatGPT does. I'm okay with that.

I create AI assistants with different skills and yes, even, personas—it's fun and as humans we're wired to personify and connect.

But fundamentally, I'm okay with neutrality in personality with AI. In the Star Trek canon, "Computer" as a character is drier than Wonder Bread run twice through the toaster. But it's beloved—a deadpan, fact-based resource available to all. That neutrality is part of what makes it trusted.

Claude can be very good at taking direction on voice and style, as I've done in the Projects feature in Claude.

This trait makes it incredibly effective as a partner for writing and editing, whether it be to pick up on your voice or writing style. It's a high leverage option if you need help drafting social captions or like me, outlining an article.

If you or your team do a lot of writing and editing, Claude is a no-brainer and outshines ChatGPT.

3. The Trust Factor

Anthropic is doing a decent job of positioning itself as the less cringe alternative to OpenAI and Grok. Now, generally speaking, LLMs are created in the same relative manner and function in the same way.

However, each company trains their model differently, but differently enough that we've noticed they take on their own style or, for lack of a better term, their own personality.

Anthropic has positioned itself as being the goody two-shoes of the AI world. They're very overt about their ethics and privacy. They hire philosophers to evaluate and train the AI to be more ethical. They publish really good education materials on how to use their AI and share research from experiments, even those with the not so great results (read what happened when Claude ran a vending machine at Anthropic and the other when it tried again at the Wall Street Journal).

But this smaller and mightier AI company is beloved, especially by coders and developers, because of its strength for technical work. But it's quickly gained a following with non-technical people in non-technical use cases, like me.

Branding matters. Marketing is a big deal.

Most of the company's revenue comes from business accounts that rely on privacy, safety, and security.

AI has an enormous PR problem. But Anthropic has positioned itself well in part due to its "human-centered" marketing (they've held in-person Claude pop-ups all over the world) and the performance to back it up.

A lot of your work and creativity requires visual assets.

It's very good at reading images, documents, et cetera, but not so great at generating graphics or images. It's not a visual model. And don't even think of asking for video.

If your work requires a lot of visual assets, graphics, and images, Google Gemini may be the better option for you with its Nano Banana model.

You'll hit usage caps especially on the free tier.

Depending on how verbose you are, how complex your work, and what plan you're on, you'll be timed out before you can use Claude again. These caps may get you frustrated enough that you don't wanna work with Anthropic, which is the primary reason why I know many people don't use Claude as their main driver.

My understanding is that enterprise teams have less of an issue with this, so it's more on individual account holders. Makes sense because Anthropic's business model is most likely to accommodate their team and enterprise users.

I suspect as compute gets cheaper, the caps will likely disappear.

In full disclosure, I pay for the $100/month tier. My usage absolutely justifies that, and I ended up removing some other subscriptions to make room in my budget for it. Even with my verbose usage—lots of editing and drafting, lots of brainstorming—I've only encountered caps twice in the past six months of heavy use.

You require the best voice mode.

As of publish date, Claude mobile app is the only place where you can voice dictate or do a voice chat with Claude directly. For some reason, those features are not available in the web app or the desktop app. And honestly, this is the biggest downside for me as I tend to speak and talk more clearly than I can type (hello, ADHD).

The mobile app's voice dictation (microphone icon) is just ok.

The mobile app's voice mode is not great. In fact, it's the weakest out of ChatGPT and Gemini. I frequently am interrupted mid-sentence or it doesn't completely catch what I'm saying.

My workaround: I brainstorm first in ChatGPT voice—it's my go-to for initial run-throughs. I get my best ideas out walking, driving, etc. So I open the ChatGPT app, click on the soundwave icon and start talking. Once I've reached a point where I've brain-dumped enough and have a clearer picture of what I want to work through, I ask ChatGPT to create a synopsis. I copy and paste that into Claude.

This also preserves my Claude capacity, which matters when usage caps are tight.

The main reason that I know most people don't like to switch, even if they're Claude-curious, is that ChatGPT knows them so well.

And that's all in part to the memory feature. ChatGPT was the first to give it to the world, and it makes interacting with it so easy and so seamless. That's what makes it feel like AI "knows" you.

You don't have to prompt it each time about your name, what you do, what you talked about last week. It remembers your preferences, your patterns, how you like things done. And that context is valuable. And that's what makes ChatGPT so sticky, why folks always want to come back to it. When Claude introduced memory in 2025 is when I made the switch from being primarily a ChatGPT daily user to a Claude stan.

And theoretically, yes, you'd have to rebuild context with Claude. But there is a way to jumpstart that...

What you need is data about you from ChatGPT to give to Claude.

Part 1: Export your data from ChatGPT.

Part 2: Ask ChatGPT to create a document you can use to introduce yourself to Claude. [GET THE PROMPTS AND PROCESS HERE]

What you're doing is essentially asking ChatGPT to talk about you. Your patterns. What it's explored with you. How you think, maybe how you like to work, maybe how you speak. Another fascinating thing to try even if you're not switching AI anytime soon.

Part 3: Take the summary that ChatGPT created for you in Part 2 and the data export and paste it into a new chat with Claude.

Instead of a blank canvas, Claude now has context of you, a new user.

There's more to AI than just ChatGPT. You have options if you're over it or you want to add on to your AI toolkit.

Exploring Claude is understanding that not all AI is created alike. Claude in many ways is a strong partner, especially for thinkers like us.

If you've made it this far, go forth and Claude. And if it changes how you think and work with AI, let me know.