Don’t Build Your House on Someone Else’s Land

With all these AI tools built off the back of OpenAI, it makes you wonder how sustainable they are.

Don’t Build Your House on Someone Else’s Land

Hey Friends 👋,

A quick note before I dig into this week’s newsletter.

I have covered a lot of varied topics in the last few months. Going forward, this newsletter will be more focused on helping developers excel in their engineering careers.

If you are not here for that, don’t worry, there will still be plenty for you too! I will still be sharing interesting picks of the week as well as links to my other writing.

I plan to make my second brain/digital garden public in the coming weeks as a way for me to write about various topics without worrying about turning away subscribers who are here for engineering content.

More on that soon! I hope you are having a good Sunday, now back to today’s post.

As the age-old proverb says, “Don't build your house on someone else's land.”

While it may seem obvious, many entrepreneurs and businesses fail to heed this advice. They build their products on top of other companies' platforms and APIs, hoping to ride the wave of their success.

However, as recent events have shown, this approach can be risky at best and devastating at worst.

What does “Building Your House on Someone Else's Land” mean?

At its core, “building your house on someone else's land” means that you're relying on someone else to provide the foundation for your business.

In other words, you're building on top of someone else's platform or API.

This can be a great way to get started quickly, as you don't need to build everything from scratch.

However, it also means that you're at the mercy of that company. They can change their rules, and pricing, or even shut down their platform altogether.

The Dangers of Building on Someone Else's Land

For those trying to grow a massive Twitter following, many have turned to the many Twitter apps out there that claim to help you grow your account.

Most of these applications have been created by indie developers using the Twitter API.

Elon Musk recently announced a change to the pricing of their Twitter API. The API used to be free, but now there are 3 tiers:

  1. Free – Low rate limit, 1,500 tweets a month.
  2. Basic ($100 per month) – Low rate limit, 3,000 tweets per user, 50,000 tweets per application.
  3. Enterprise – anything above basic

For these Twitter applications that have a decent number of users, the basic tier doesn’t cut it.

The Enterprise tier however starts at $42,000 a month!

There is no tier between Basic and Enterprise. You either have a small application that can serve a few users, or you have half a million bills each year.

For example, the following businesses have either been forced to shut down or require users to provide their own Twitter API key.

In the last few weeks, thousands of AI businesses have all spun up. All of them are using OpenAI APIs, which at the moment are incredibly cheap.

Some of the most popular AI writing tools such as and the AI built into Notion are all using OpenAI under the hood.

Currently, OpenAI has the following pricing for their API:

  • gpt-3.5-turbo: $0.002 / 1K tokens
  • gpt-4: $0.03 - $0.12 / 1K tokens (differs based on prompt/completion and 8K/32K context)

At the moment, it is significantly cheaper to use the OpenAI API than it is to pay for ChatGPT Plus. On top of that, you only pay for what you use.

Unless you are generating around 100,000 words a month, the API is a better option.

You can also use the API key with tools such as TypingMind and Auto-GPT.

With all these companies building on top of OpenAI, what is going to happen if they increase their prices drastically, as Twitter did?

What is the alternative?

If you are out to make a quick buck, then of course using someone else’s platform is one way you can do that.

Many of the Twitter tools managed to make millions before Elon increased the price. If OpenAI doesn’t change its pricing, then it certainly is cheaper than running your own AI model in the cloud.

However, if your business is simply an API call to another platform, then what moat are you creating for your company?

What is to stop a competitor from coming along and doing exactly the same thing?

Many of these companies that are simply a wrapper around OpenAI are likely to find themselves replaced with ChatGPT plugins in the near future.

Where you can, you should always try to build products that don’t rely on other platforms to work. Obviously, there is a limit to this. I am not suggesting that you shouldn’t use AWS or Azure and host your own infrastructure.

Don’t build your core business around someone else’s product that you can’t host yourself or there aren’t other options available.

Luckily, on the AI side, there are a few alternatives that can be used commercially which can be hosted yourself.

  • GPT4All - This is a complete open-source model that has recently switched to using GPT-J instead of Llama, so can now be used commercially. It also comes with a simple installer for Windows, Mac and Linux, so you can now run your own ChatGPT-like chatbot for free.
  • Dolly v2 - This is another open-source model that can also be used commercially.

I am looking forward to these open-source models becoming more advanced in the future so that OpenAI doesn’t have as much of a monopoly as it does at the moment.

❤️ Picks of the Week

📥 Newsletter - How I Write and Grow Refactoring in 2023 ✏️. I think writing is incredibly important for software developers, and one of the best ways is to write in public with a blog or a newsletter. Luca who runs the Refactoring newsletter recently wrote about his process for note-taking and writing which is worth a read.

🐦 Tweet - How does a Large Language Model like ChatGPT actually work? I have been trying to wrap my head around LLMs and Neural Networks, and this tweet explains it so simply. For anyone who wants to understand them, this is worth a read.

📝 Article - What Is ChatGPT Doing … and Why Does It Work? If you want the in-depth version, then this is it. Written by Stephen Wolfram, who is the creator of Wolfram|Alpha. I have started reading this, but I definitely need to take some notes.

📥 Newsletter - What does Gen Z really think about work? I saw this shared on Substack Notes, it's an interesting view on how people’s relationship with work is changing. Especially given the example is from a fellow software engineer.

👨‍💻 Latest from me

🎬 YouTube - 10 Questions You MUST Ask in Your Next Software Engineering Interview. I am always surprised at the end of conducting the interview when the candidate hasn't got any questions for me. Surely if they were interested in the position, they would want to ask me something. In a lot of cases, our minds just go blank due to the stress of the situation. I am hoping this video will help those who struggle with things to ask.

💬 Quote of the Week

The final form of leverage is brand new—the most democratic form. It is: “products with no marginal cost of replication.” This includes books, media, movies, and code. Code is probably the most powerful form of permissionless leverage. All you need is a computer - you don’t need anyone’s permission.

From The Almanack of Naval Ravikant (affiliate link) by Eric Jorgenson. Resurfaced with Readwise.