Don't learn programming - learn how to plant potato

Last week, I created three MacOS programs:

1. downloads all transactions from my stock broker – parses them and generates a tax report;
2. generates text (subtitles) based on what’s said in a video;
3. downloads all available financial data about specified stocks.

I’m not a programmer, and I’m not writing a brag post. See, recently, NVIDIA’s CEO, Jensen Huang stated that by 2027, anyone will be able to create programs in plain conversational language thanks to artificial intelligence. So, it’s better to teach kids how to plant potatoes than to teach them programming languages.

When such a respected person says something like this, you’ve got to believe it. But I decided to test it. I used the free version (without a subscription) of ChatGPT. But what should I ask it to do?

First, I decided to make my tax reporting easier. Last year, I accumulated about 500 transactions that I needed to analyze painstakingly in Excel. I described the task to my "programmer," provided the input and output file formats – and voilà! Within a couple of hours, I had a Proof of Concept (PoC) written in Xcode – with a terrible interface and a primitive algorithm, but hey, every great journey starts somewhere.

Then I thought, since my friend and I make a podcast, it would be great to generate subtitles in SRT format (with word-by-word timestamps). A few more hours later, I already had a PoC that uploads a video or audio file, detects the language spoken (if there is one), and generates text.

This reminded me of an old idea that came to me during a finance class at IMD. There, we were taught which financial indicators and ratios to look at to determine whether a company is a good business.

I thought back then: Damn, all these indicators are available online, but it’s inconvenient to analyze them manually, company by company. What if I create a bot that goes through a list of stocks, collects this data for me, and then I analyze it in a convenient table to build my optimal investment portfolio?

And so, in two days, ChatGPT and I, if you can say that, wrote such a program – and it works. I won’t bother you with technical details, but even if I had plenty of free time (nervous laughter), it would have taken me months to fine-tune a mechanism that browses the internet, mimics different browsers and devices, parses the received HTML, and saves the necessary data in the required format.

Disclaimer: I’m not writing programs for sale or doing anything illegal – I’m just boosting my productivity.

Some observations:

1. You can’t say Jensen Huang’s prophecy has come true just yet – you still need to understand code and the basics of object-oriented programming to communicate with AI in the same language. But I do not doubt that by 2027, or even sooner, it will become a reality. A big hello to all programmers out there!
2. ChatGPT doesn’t write code from start to finish, at least not in the free version. The interaction feels a bit like dealing with a large corporation’s support service, where every time you’re transferred to the next expert, you have to explain everything from scratch. But since I’m a systems engineer, I’m used to breaking down big tasks into smaller ones.
3. ChatGPT can already refactor your code – this is when you’ve written everything in one huge, unreadable file and then decide it’s time to organize it. For those unfamiliar with programming – it’s like saving all your files to your desktop and one day deciding to sort them into folders.
4. This is already a fantastic tool for anyone who needs to create an MVP or PoC but doesn’t have the budget for developers. ChatGPT can write in any language and for any platform, even VBA in Excel on GamePad.
5. ChatGPT is an excellent debugging partner. Now, finding and fixing those annoying little code errors is much easier. Moreover, it can explain any confusing lines of code. Productivity is through the roof.

While writing this, my program finished downloading data for the S&P 500, and now I’m off to read about how to plant potatoes.