GitHub Copilot: How we created an app with speed and agility

We put the new GitHub Copilot to the test. Here are the
pros and cons and whether it is the right fit for your project. 

GitHub Copilot is one of the latest AI coding assistants promising jaw dropping apps in record time. We put it to the test with a simple task (or prompt): create a web app to manage shared office desks. Here’s what we learned along the way.

What is GitHub Copilot?

GitHub Copilot is an AI assistant, powered by OpenAI’s Codex model. It integrates directly into your development environment (mainly through a Visual Studio Code extension) and suggests code based on the context of your project.

What makes Copilot stand out is that it understands your code and helps you write it in real time. As you type, Copilot predicts what you might need next, from simple syntax fixes to more complex functions. It’s like having multiple tools combined into one: it acts as an autocomplete, provides auto-commenting for your code, generates tests, suggests fixes for errors, and, of course, serves as an intelligent AI assistant ready to lend a hand at just the right moment. Why did we choose GitHub Copilot over other AI assistants?

There are a lot of AI coding assistants out there, but GitHub Copilot has a few features that stand out:

• It’s backed by GitHub, one of the most trusted names in the coding world.
• The integration with Visual Studio Code (VSCode) is seamless. It feels like Copilot was made to live right there with your code, making it incredibly easy to use and add suggestions as you go.
• It saves a lot of time and reduces mistakes, helping you focus on writing better code faster.
• It’s affordable, with one of the lowest price points on the market, and even offers a free trial to get started.

We chose GitHub Copilot because it’s reliable, integrates effortlessly, and delivers great performance without breaking the bank.

Building a web app at the speed of light with GitHub Copilot

We wanted to see how quickly we could go from idea to reality, all while keeping it functional, sleek, and ready to deploy. The goal? Build a simple web app for managing office desks at lightning speed and host it using GitHub Azure services.

The app needed to display a list of desks, make them selectable, and be user-friendly. Nothing too crazy, but we wanted to get the essentials down without spending hours tweaking every little detail. We needed something that worked – and worked fast – without sacrificing quality. Armed with a detailed prompt, we jumped in.

Our first request was straightforward: generate a backend with Django and a frontend with React. It generated a basic (but functional) list, exactly as requested. A far cry from the polished app we had in mind, but definitely a solid starting point.

From here, we took a hands-on approach – analyzing each piece of code Copilot generated, checking for quality, and fine-tuning things as we went. If something wasn’t quite right, we would tweak the prompt and ask for a revision. We kept things moving quickly, making small, manageable changes one at a time, and seeing how Copilot responded to each new request.

While Copilot is excellent at churning out backend code and handling the heavy lifting, it still struggles when it comes to the more nuanced aspects of UI/UX. For example, we asked it to display items in a clean, responsive layout, with some extra flair – like swapping text for icons or arranging items in neat rows. While the results were functional, they lacked the polish and attention to detail we were looking for.

This is where the developer’s role became critical. Copilot did an amazing job getting us started, speeding up the process, and providing the foundation we needed. But the real magic happened when we stepped in to refine the app, focusing on usability, aesthetics, and those finishing touches that make a product truly stand out. Copilot accelerated our workflow, but it was the human intervention that turned a functional prototype into something we’re truly proud of.

The good: what GitHub Copilot did right

Copilot delivered in these key areas.

• Detailed responses: GitHub Copilot didn’t just provide the code, it explained it. Step-by-step breakdowns, making it easier to understand the logic behind the code.
• Copy-paste magic: Integrating generated code into your project is absurdly easy. Just click, and it’s in there. No missed lines.
• Rapid development: From concept to functional app in under 30 minutes? Yes, please. GitHub Copilot was like our fast-track ticket to a working web app.
• SQL genius: Ask it to generate a SQL script to populate a test database, and you won’t need to lift a finger. It just works.

The not-so good: GitHub Copilot’s flaws

As much as GitHub Copilot impressed us, there were definitely a few rough edges that we couldn’t ignore.

• Handling complex prompts: When the prompt gets too long or complicated, GitHub Copilot can get a little lost. It sometimes misses parts of the requirement and gives partial answers.
• Memory lapses: This is where things get quirky. GitHub Copilot’s memory doesn’t always stick. Tell it something didn’t work, and it might just repeat the same answer like it never heard you. It also forgets context after a while, which can be… frustrating.

Verdict: would we use it again?

Yes, absolutely. But with a caveat. GitHub Copilot is not a replacement for a developer. It’s a tool, not a mind. It excels at reducing the grunt work – generating boilerplate code, refactoring, and handling those wearying details you’d rather not waste time on.

That said, there are still some areas for improvement – namely, handling more complex prompts, improving memory retention, and refining its approach to UI/UX tasks. But given the progress we’ve seen, we’re confident that future iterations will only get better. So, it’s still a work in progress. A very capable work in progress.

And if you’re looking to accelerate your development process, solve complex problems, or just want a reliable partner to bring your ideas to life, let’s talk. Near Partner has the tools, experience, and innovative mindset to turn your insights into execution. Reach out, and see what we can create when we work as a team.