Ep. 33: ChatGPT vs. Claude - Real Differences After Weeks of Daily Use

LISTEN TO THE EPISODE

In this episode I share my personal experience of what it’s actually been like to ditch ChatGPT and go all-in on Claude. This episode covers the real differences between the two platforms: what Claude does better, where it falls short, and why the switch probably isn’t as big a deal as some people are making it out to be.

First, Some Context: Anthropic Is Putting Up the Good Fight

So, worth noting: At the time of write this, Anthropic is scoring some serious brownie points with US-based users. Anthropic has refused to remove safeguards that would prevent its technology from being used to target weapons autonomously and conduct U.S. domestic surveillance.

The government gave Anthropic until 5pm on Friday, February 27th, to comply with their demands for unrestricted military use or risk being labeled a supply chain risk. Anthropic said NOPE, and Trump threw a temper tantrum.

What this means for Anthropic going forward is TBD, but they’re definitely winning over the US people. Sam Altman, likely spurred by seeing folks leaving ChatGPT to head to Claude, actually came out with a statement saying he shares the “red lines” set by Anthropic, restricting how the military uses AI models.

Anthropic is still in bed with Palantir, so they’re no angels, but this is definitely a win.

We’ll see what unfolds in the upcoming weeks.

About Making the Switch: It’s Not That Serious

I’ve been using Claude exclusively for the past few weeks, and because I’m online way too much I’ve continued to see folks on Threads making a big to-do about switching from ChatGPT to Claude. Insert massive eye-roll. So I figured I’d throw my hat in the ring.

First off: it’s not that serious. I keep seeing people hemming and hawing about the best way to bring their info over to Claude. Insert another massive eye-roll. Here’s my take:

1. Read Episode 30. I outlined exactly how to make the switch in the most painless way possible.
2. In my humble opinion, if someone is asking on Threads how to make the switch (why are we crowd-sourcing this?) it tells me they don’t use ChatGPT nearly enough to have nearly enough info on there to actually worry about making said switch. High volume users with enough history/projects/custom GPTs/usage to actually fret about said switch would know they could just ask ChatGPT how to do it.
3. Again, it’s not that serious. Just start using Claude. You’ll be fine.

My Experience So Far

The two LLMs absolutely have different “personalities”, but for what most of us use it for, it’s fine, and they both get the job done.

Here, however, are the biggest differences I’ve noticed:

1. Claude doesn’t have a native image generator.

Honestly? Good for them. Any AI-image slop on the internet, we now automatically know it’s not from Anthropic. For those needing an image (I make them for my blog post featured images), the solution is to use Google’s image generator, Nano Banana.. And yeah, I know I said in a previous episode that if forced to pick an AI race winner it would be Google, but JFC Gemini is slow and I don’t like it. The image generator, though? It’s good.

2. Claude has usage limits.

I spoke about this two episodes ago in ep. 31: Are We Addicted to AI, and I continue to be mindful of it. Realistically, for the average user and general use, I don’t think you’ll come close to hitting your limits. If you start generating code, that’s a different story, and namely, in my experience, with session usage limits, not weekly usage limits.

3. Claude doesn’t have Custom GPTs.

No matter what folks on the internet say, Claude doesn’t have Custom GPTs. Claude has Projects (and also Skills), neither of which are Custom GPTs. Claude’s Projects are the same as ChatGPT’s Projects, which is really nice, but neither of those two things are Custom GPTs. Admittedly, this is a big loss, particularly for folks like me who shared Custom GPTs with clients as a resource.

4. Claude has a tendency to break.

I’m a night owl, and more than once, always during evening sessions, Claude has just stopped working mid-session. It gives a little “retry” message, re-tries 10 times, and then gives up. I’m not sure what causes it, and it’s by no means all the time, but it’s definitely happened way more times than I ever experienced when using ChatGPT.

5. Claude’s persistent memory doesn’t seem to be as good.

I’m pretty sure the simple solution here is to use Projects and upload documents with the information that you want it to remember, but I do find that Claude needs more reminding about things than ChatGPT did. Not a deal-breaker, but worth noting.

6. Claude doesn’t have voice mode.

Voice mode is where the AI actually speaks back to you. Some ChatGPT users found it weird; I actually liked it. Overall, I don’t think it’s a bad thing that Claude doesn’t have it. Voice mode, among other things, registered to me as part of OpenAI’s attempts to personify their product and get people addicted. No bueno. Again, I liked it, and I’m not upset that Claude doesn’t have it, but it is a difference worth noting.

What Claude Does Better

On a more positive note: I find Claude to be less verbose than ChatGPT, and I really like that. ChatGPT is chatty AF, and despite me updating my custom instructions, it always defaulted to saying too damn much. I do think this speaks to a real, fundamental difference between the two LLMs, because I use the same custom instructions in Claude that I used in ChatGPT, and Claude knows when to shut up.

Personality-wise, I find Claude to be more objective and less sycophantic than ChatGPT. I think that part of what makes it less sycophantic is its default to ask clarifying questions, which, admittedly, might be a product of my custom instructions. I need to go and check. Either way, Claude definitely does not seem nearly as concerned with being your yes man so that you keep using it.

On writing: I can’t say I prefer one over the other because I’ve never had either of them just fully write something for me. Yes, I use them to write this Curious Companion, but it generates the Companion from such a detailed outline and adds so little, it’s not a great assessment. I know some folks prefer Claude for writing, but I’m gonna encourage you to keep using your brains and writing things yourself.

How I Used Claude This Week

Each episode I share a quick example of how I used Claude that week.

This week I used it for one of my favorite things: writing a script for Google Apps Script. Google Apps Script is a mini-program you can build inside Google’s tools, like Sheets, Docs, or Gmail, that automates tasks for you so you don’t have to do them manually.

Previously, I’d used Google Apps Script (with the help of ChatGPT) to set up an automation that let me email resources (YouTube links, article links, etc) to myself and have them automatically post to my website.

This time, I had Claude help me with my scheduling software, Acuity. Acuity isn’t Google-based, but via the API (application programming interface), I can now have all appointments automatically populate into a Google Sheet. Why? Because you cannot natively sort contacts by appointment type inside of Acuity. So dumb.

I hate Zapier. Used it once, it broke, I never forgave it. (Also I hated the idea of paying for something that could break so easily.) So I take every opportunity I can to vibe code free solutions, and this was one of them.

Full transparency: Apps Script is not the most user-friendly, but that’s the beauty of using AI. Not only will it write the script, it’ll tell you every single step you need to take to get it running.

Da Wrap-up

Overall, the switch to Claude has been totally fine. I don’t find one to be drastically better or worse than the other. If you’re wanting to switch but you’re on the fence, just do it. You’ll be fine.

As always, endlessly appreciative for you and your curiosity.

Catch you next Thursday.

Maestro out.