A recent trip to the US took me to The Golden Triangle. In case you are, wondering, this describes LA’s upscale retail district centred around Rodeo Drive; but so-called because the district is bounded by a triangle of roads (Wilshire, Canon and Santa Monica, for the curious).
While I knew about the retail district before I arrived in LA, I had never come across the ‘Golden Triangle’ label before.
But it brought a smile to my face because just before finding out about LA’s Golden Triangle, I realised I’d been thinking and learning about another ‘golden triangle’ – this time related to AI and custom GPTs.
My AI-related ‘golden triangle’ consists of three different ‘languages’: Markdown, JSON and Python.
I hadn’t given this clutch of technical terminology the ‘golden triangle’ label before my arrival in LA; but when I heard reference to the retail Golden Triangle, it seemed like the perfect label for the tech stuff, too.
Just as Wilshire, Canon and Santa Monica connect to each other to create the boundaries of a greater whole, these three languages all connect to each other to help a custom GPT work more effectively and efficiently.
If you’ve read any of my previous posts, you will know about my experience with using Information Mapping. It was this experience and an earlier conversation with Scott Schang (read about that here) that led me to understanding my golden triangle.
It started with Markdown. I’d heard the term in relation to AI and custom GPTs but didn’t really understand what it was and what it did. After talking about this a little with Scott at lunch, a few hours later I found myself in conversation with ChatGPT to dig into the detail.
Thankfully, it’s not that complicated. Markdown is, essentially, a simple way to tell your GPT how to format and present text to the GPT’s user, when responding to some input from the user.
You set up a series of formatting ‘rules’ and instruct the GPT to apply the rules consistently to any text it is generating. This ensures that text is well laid out and easy to read. If you don’t provide any rules around this, the formatting of output tends to be a bit hit and miss.
Markdown is very easy to understand and set (I’ll do a more detailed post on this soon); and if you know anything about HTML, you will see the similarity immediately.
For me, the icing on this particular cake, is the connection to Information Mapping. There is a great deal of flexibility in how you set up Markdown. This means it is relatively easy to mimic the hierarchy and presentation styles of the Information Mapping approach.
So, in summary, Markdown is about how content and data is formatted and presented to the users of a given GPT.
This led me to ask ChatGPT about JSON. Another term I had heard bandied about a lot; but one that I didn’t fully understand yet.
JSON actually stands for JavaScript Object Notation. Like so much about tech speak it can sound intimidating before you’ve even begun. But actually, the scary-sounding name conceals a very simple idea. JSON is a way of structuring data that is easy to read for both humans and machines. It can also be used in almost every programming language.
This last point is hugely important. Because it means that JSON makes it easy for different systems, apps and languages to ‘talk’ to each other, i.e., to pass data to each other without breaking.
So, in summary, JSON is about structuring data and moving it between different tools, apps and databases.
Finally, I was curious about Python, a programming language that seems to get a lot of attention in the world of AI and custom GPTs. I’m certainly not going to claim that Python is easy to write or understand. (The writing of it is definitely something I’d outsource to AI!) But its purpose is relatively easy to understand.
In essence, you can use Python to do stuff with your data. Too long a list to cover here but some simple things might include: looking up and fetching data from a database or carrying out a quick calculation.
So, that’s my golden triangle. JSON: a way to structure data and make it easy to pass between systems and apps. Python: a means to work on that structured data. Markdown: a way to control how data is presented to users.
I’ll write a bit more detail about each of the parts of my golden triangle in the coming weeks.
Stay tuned!
