The Rules of OpenAI (And Why They Hurt You)

In the early stages of building Juice, we had a choice to make: whether we would use GPT-3  (an advanced NLP model created by OpenAI) or comparable open source technology.

I want to make it clear from the outset that we appreciate the numerous innovations that OpenAI has minted over the years. They are in part responsible for the current AI renaissance, both technologically and culturally.

However, after researching OpenAI and its policies, we realized that building on OAI was simply not an option in order to bring you the best possible products.

And I bet the reasons why will surprise you.

Why open source matters

We made Juice to empower youthe creator.

Maybe you're an entrepreneur with a brilliant new product, or the head of SEM at a boutique marketing agency.

No matter what you create, we believe that you deserve access to high-caliber marketing tech that allows you to compete against much larger businesses.

And therein lies the problem – while competition is good for everyone, OpenAI's policies have become increasingly anticompetitive. You see, OpenAI doesn't necessarily want you to be able to create things.

How OpenAI became ClosedAI

A few years ago, OpenAI licensed GPT-3 exclusively to Microsoft.

This means that any companies built on GPT-3 must abide by OpenAI/Microsoft's usage guidelines or risk losing access to GPT-3 entirely.

Not only that, but the more we looked into it, the more we started seeing all sorts of red flags with how OpenAI allowed people to use their technology.

Red flag #1: GPT-3 LIMITS the amount of words you can generate (it can't output 1,500 word articles)

No matter how advanced GPT-3 gets, you’ll never be able to generate more than a few sentences at a time with a single click.

In fact, OpenAI’s usage guidelines state that NO company can use GPT-3 to output full-length articles.

That’s right – OpenAI explicitly limits the number of words users can output with a single click (the limit is 250 ‘tokens’).

In AI-speak, ‘tokens’ and number of words are directly correlated. 250 tokens is about 80-120 words, on average.

To put into perspective how limiting this is, let’s compare it to Juice.

Every time you generate an article with Juice, you’re getting upwards of 4,000 output tokens, which is closer to 1,500 words.

Why can Juice instantly output full 1,700-word articles, whereas GPT-3 products top out at 120 word paragraphs?

The answer is simple: Juice isn’t built on GPT-3.

Red flag #2: GPT-3 censors all of your content before you even see it

We’re not making this one up either.

Similar to how OpenAI limits the number of tokens GPT-3 can output, so too does it filter everything GPT-3 products output before it reaches your screen.

Juice doesn’t censor anything, unless you tell it to. You’re an adult, so we treat you like one.

No matter your industry or SEO goals, Juice entrusts you to make decisions that will help you succeed.

Juice allows you to create your way

We ran an SEO agency for years, creating the ultimate formula that allows smaller businesses to take advantage of search to get traffic and boost sales.

Using truly open source AI is the only way to make this happen.

For more information, here is a link to our documentation explaining what Juice is and why you should use it!

About The Author

Derek Sturman
Co-founder of, CEO of Panda
Juice Beta is ending July 1st! Subscribe before end of month to lock in Juice Plus for 50% off!
$49 $25
Sign up now
Juice Beta is ending soon! Subscribe now to lock in Juice Plus for $49 $25
Sign up now