Two weeks ago, I found myself leading a unique project. We enlisted the talents of two programmers to develop a Minimum Viable Product (MVP). Both of these professionals had previously worked with me, making this a reunion of sorts.

First, we have Alex from Germany, a seasoned coder with a staggering 19 years of experience. His approach to programming was purely traditional; he worked solely with code.

On the other hand, we had Hamid from Pakistan, with a mere four years of professional experience under his belt. However, Hamid employed a mix of traditional coding, GitHub’s Copilot, OpenAI’s GPT-4 language model, and no-code platforms in his development process.

The project parameters were clear: both programmers were provided with Figma screens and detailed specifications. They had a designer on hand to help with the necessary assets and pre-existing code that required integration.

What transpired over the next week was, quite frankly, astonishing.

In just seven days, Hamid delivered the first version of the MVP, boasting 100% test coverage of both the coded and no-code portions. Impressively, an estimated 95% of the overall work seemed completed and operational on initial observation.

Hamid utilized @bubble for UI and front-end workflows, generated Cloudflare Workers with the help of GPT-4, incorporated existing code through Copilot, and orchestrated tests using GPT-4 (playwright/ava).

In terms of cost breakdown, Hamid’s work required:

  • GPT-4: $211
  • Copilot: $20
  • Cloudflare: $5
  • Bubble: $134
  • Compensation: $2460 (41 hours of work)

The ongoing hosting and operational cost were estimated at $139 per month.

In stark contrast, Alex managed to complete roughly 7% of the assigned tasks within the same timeframe. His costs involved:

  • Vercel: $20
  • Compensation: $3500

Alex’s estimation for completing the entire project was a significant $45k, with an additional $11k earmarked for testing. His hosting and running costs were projected to be a mere $20 per month.

The contrast between the two developers was stark, and the results were unexpected. This project, initially taken up to satisfy personal curiosity, evolved into an enlightening experiment for both myself and my friend, who owns a development agency. Our initial estimates had Alex taking slightly longer than Hamid, but we never anticipated this degree of disparity.

When we discussed the results with Alex, his defense was, “But it will be so much cheaper to run this app, and you’ll have everything under control.” He seemed oblivious to the opportunity cost of a slower delivery time and a significantly higher development expense.

In the end, we had to part ways with Alex, primarily because he was rigid in his development methodology, showing little faith in no-code/AI tools.

This experience made us realize that the future might be leaning more towards versatile developers like Hamid. My friend’s development agency, home to over a hundred developers like Alex, is now considering a significant shift in their training approach or even replacing their staff.

It seems clear that five years from now, developers who are adaptable and open to the benefits of AI and no-code will still be in high demand, while traditional coders may need to reconsider their career trajectories.

What’s your perspective on this? How do you see the software development landscape changing? Let us know in the comments below.

This blog post was inspired by a discussion on a Twitter thread. You can find the original conversation

Ab Advany

Leave A Comment