Hey everyone,

Today is the first weekly deep dive, the second format to the Sunday Space.

For those who missed this week’s Sunday Space, click here.

We’ve got a lot to cover today, so let’s get right to it.

Today at a Glance

• This week, OpenAI launched its highly anticipated GPT Store — which aggregates and categorises “GPTs” created by the developer community for public use.

• GPTs are AI Assistants — they’re custom versions of ChatGPT trained by the developer for a specific use case.

• The GPT Store is being touted as the Apple App Store equivalent for AI. I’m not so sure about that, given the extremely low barrier to entry when it comes to creating GPTs.

AI's App Store

This week, OpenAI launched the highly-anticipated GPT Store to the public.

It’s a moment that the AI developer and indie hacker community have been waiting for since OpenAI’s Dev Day in November 2023.

It’s worth crediting Sam Altman and the OpenAI team for reaching this milestone. The rate at which they’ve shipped products and innovated since ChatGPT’s launch has been something to behold.

Even with Altman’s ouster from the CEO role just days after Dev Day and the following madness, they’ve still managed to deliver.

Incredible.

There are two main reasons for the intense hype around the GPT Store:

• GPTs are, in a way, AI “apps” built on GPT-4, the AI model powering ChatGPT.

• GPT Builders can participate in a shared-revenue program with OpenAI and be compensated for the apps they create and sell on the platform.

The similarities to the Apple App Store are obvious — developers across the globe can build applications permissionlessly and distribute their products to millions of users.

Naturally, the AI community has looked back to history to extrapolate how the GPT Store will turn out based on the App Store.

And rightly so — look no further than Uber, Airbnb and Doordash, among many others. These are incredible businesses, but they were enabled by the App Store and turned into multi-billion dollar companies because of it.

While it is true that studying history closely can inform the future, a quote from Mark Twain adds the necessary nuance to this notion:

So before I explain why I don’t think the GPT Store will play out exactly like the App Store (and offend all those thinking they’ll make millions from it), let’s step back to understand the concept of what GPTs exactly are and how they work.

GPTs 101

GPTs are custom versions of ChatGPT created for a specific purpose.

They’re intended to help users make ChatGPT be more helpful with their daily tasks — from meal planning to email writing to coding.

GPTs created through the ChatGPT interface are insanely easy to make and share with friends or work colleagues through links. Publishing one’s GPT to “Everyone” hosts it on the GPT Store.

They can be created by anyone—no coding knowledge is needed.

Creating one is as easy as starting a conversation with the “GPT Builder” interface and crafting your GPT’s instructions, personality and capabilities in natural language.

Once the foundations are set (which can take ~5 minutes), GPTs can be as simple or as complicated as users decide to engineer them.

Users can upload files to their GPTs to be used as knowledge or define custom actions by making APIs available to the GPT.

This is crucial because it allows the GPT to connect to external and potentially proprietary data sources. Think about connecting it to external databases or plugging it into emails. This ability is what will separate the best GPTs from the rest. More on that below.

For detailed instructions on how to create your own GPTs, click here.

My GPTs

Since OpenAI launched GPTs, I’ve spent a bunch of time experimenting with their tools and creating my own GPTs. Check them out below:

• Brian Chesky’s CTO — Coding assistant to help build marketplaces.

• TheraGPT — A therapist in your pocket, best used with voice chat.

• Silicon Scribe — Your late-night tech writing mentor and editor.

• FRS 102 Helper — A niche one for UK auditors and accountants.

And the one I’m most proud of:

• Budgeting Bestie — An assistant built on my own personal budgeting spreadsheet that I’ve used and continuously updated for the past five years.

If you’re brave enough to give them a whirl, let me know how it goes.

Doing The Hard Thing Wins

An unpopular opinion (that’s maybe less unpopular than I thought) is that the GPT Store isn’t going to be as big of a deal as most people think.

The reality is that the overwhelming majority of GPTs (99.999% of them) will be useless — they’ll be no more valuable than using optimised prompts with ChatGPT.

The few GPTs that rise to the top (the ones that people might pay for) will be those with access to proprietary data not available in the base model and those that create custom actions that give it deeper functionality outside of ChatGPT.

In the same way that data is the fuel for training powerful AI models, proprietary data that GPT-4 hasn’t already been trained on will form the competitive moat for the best GPTs.

The above will be obvious if you spend some time testing the GPTs I linked above and if you try creating your own.

This is why I believe the GPT Store will not be as lucrative for developers as the Apple App Store has been.

As a quick illustration, OpenAI announced over 3M GPTs have already been created. There are 1.8M apps available on the Apple App Store today.
Let that sink in.

OpenAI hasn’t announced its revenue program for GPTs yet, but Apple has set the bar for performant ecosystems.

App Store developers generated $1.1 trillion in Revenue in 2022. This highlights the extreme utility that apps bring society, given consumers are willing to part ways with so much of their cash to use them.

I’m not writing OpenAI off (I think they’ll be a trillion-dollar company like Apple), but it will be a long journey for their ecosystem to reach these levels:

This debate is not about OpenAI and Apple’s abilities to create revolutionary platforms. The comparison shows that to create extreme wealth in society, there needs to be an added difficulty level to the task at hand.

For GPTs, there are zero barriers to entry. I can create one by spending 5 minutes writing a few instructions without knowing how to code.

For Apps, there are considerably higher barriers to entry. I need to know how to code software, build a team, and create an actual business.

The supply and demand metrics that drive value clearly tilt in favour of mobile apps.

Similarly, the GPTs and custom Assistant products that are impossible for the average Joe to make themselves will create real economic value.

At the end of the day, the only developers who will get rich from the GPT Store are those who are willing to do the hard thing—spend more than 5 minutes on instructions, gather proprietary data from their customers or other sources and craft custom actions that enhance the experience.

In short, it’s about creating amazing products that your customers love.

It always has been and always will be, no matter what revolutionary technological wave humanity is going through.

Was this newsletter forwarded to you? If you liked it, sign up here.

And, if you loved it, forward it to your friends and family.

Thanks for reading today’s deep dive.

See you in the Sunday Space.

— Luca