I’m Cheating On ChatGPT And I Don’t Care Who Knows It

Read Time: 5 minutes

…I’m a two-timer.

In my University of New Hampshire classes, I recommend that my students try playing with TWO AI models:

1. Your work model (for most of us, Google Gemini or Microsoft Copilot); and

2. Your home model (ChatGPT, Claude, or Meta’s LLama)

Why?

I’m not smart enough to predict which model will make the next significant leap forward, so I want you to be prepared. 😅

This week, it was Claude 3.5 with Artifacts 🤯… but more on that in Friday’s Deep Dive. 👀

For the last year, the AI models have all been somewhat boringly similar…

…But now we’re seeing them releasing different features and taking unique paths to training the models.

The result?

Clearer differences between AI models. 👇🏼

THE DIFFERENCES ✔️

CLAUDE 3.5: Better at Code, Cheaper, 200K context window

Here’s the deal. Claude emphasizes safety and has great new features like Artifacts. It’s been getting better rapidly.

What does this mean? I’m using Claude 3.5 for multi-step prompts and learning to play with code/artifacts.

CHATGPT-4: The Best Overall (Yes, still)

ChatGPT is powered by OpenAI’s GPT architecture, which is known for its unparalleled language understanding capabilities.

What this means? GPT4.0 is able to generate the most human-like responses across a broader range of topics with better precision. It’s better and faster if you regularly use spreadsheets as a data source and with data extraction in general. The downside? The model can inadvertently learn and then reproduce problematic patterns present in its wider training data.

GEMINI: The Best For Research

Gemini is generally considered an LLM for those on the Google stack of tools (Google Sheets and Docs). With a larger context window, it can review more content or code.

What does this mean? Choose Gemini for broader research and real-time access to the internet, as Google crawls the web more often than anyone else.

My Favorite ChatGPT Alternative 🤓

If you’re ‘brand-spanking’ new to AI, I’d skip this step.

But if you have been playing with ChatGPT for a year and are looking to branch out with your ‘V2’; give Claude 3.5 a go.

STEAL MY PROMPTS 🥷🏼

The subject line of this email was written using the following prompt…

‘I want you to act as a highly skilled marketer and top-tier email newsletter copywriter who is fluent in English. Pretend that you are an expert at writing all types of email subject lines in English.

Your task is to rewrite this email subject line of 70 characters that I will give you at the end in these formats:

1. Rewrite it to make it more concise

2. Rewrite it to improve the flow and coherence

3. Rewrite it to make it more engaging

4. Rewrite it to make it more persuasive

5. Rewrite it to make it more sassy with emoji

6. Rewrite it to make it more informal

7. Rewrite it to add more detail

8. Rewrite it to remove unnecessary words

9. Rewrite it to change the tone

10. Rewrite it to make it more accurate

11. Rewrite it but focus on clarity and brevity

12. Rewrite this paragraph in simpler terms

Bold each with big title and use format word number wise before each.

Please do not echo my prompt.

Do not remind me what I asked you for.

Do not apologize.

Do not self-reference.

Take the best action you can.’

If you want more where this came from, I’ve got an extensive prompt cheatsheet in the Resource Library.

P.S If you don’t have time for two full models in your work life, try two on your phone.

ChatGPT is my go-to, but since I’m an Android user, I use Gemini on my phone a good bit.