Hey! Let’s get something out of the way: this article has not been written by an AI.
I know, it sucks. I am sorry.
These are just some very human thoughts of mine, packed with emojis and questionable grammar choices.
Thoughts about AI, though.
In fact, over the past two weeks, basically all conversations I had with peers touched on the AI subject somehow.
That’s because — for those of you who read this newsletter from inside a cave — OpenAI just released a new chat bot, called ChatGPT, that seems impossibly good at… everything.
My Twitter feed has been, and still is, inundated with ChatGPT examples. From writing essays, to coding, to navigating the browser and casually make a dinner reservation.
The truth is, I started writing down ideas for an “AI & coding” article a long time ago. Reading them back today, some are like “Copilot is here to stay”. It seems cute now, but it had to feel like a big insight back then.
“Back then” is less than one year ago.
Github Copilot was released in private beta in June 2021, and opened up to all developers after exactly one year. Copilot, in turn, is powered by Codex, which is developed by OpenAI, just like ChatGPT.
Codex is a constrained, fine tuned version of the GPT-3 language model, while ChatGPT is based on what OpenAI calls GPT-3.5, a marked improvement on the former. So it is safe to expect the same kind of improvement to come to Copilot soon.
How will these tools change the way we write code? And, in turn, how will this change our industry?
This article is a reflection on the first, second, and third order effects of AI-assisted software development, which right now feels like an inevitable future.
Here is what I will cover:
💻 Workflow — I spent two days writing a Telegram Bot with the help of ChatGPT. I got my hands dirty, and I have thoughts.
🐛 Correctness & Bug Fixing — AI rapidly changed my assumptions about correctness and my approach to fixing bugs. I wrote a reflection on where this is going.
🔄 Feedback Loop — thoughts on how these models can improve in the future, with the help of human experience. Which will have a big impact on… 👇
💡 Innovation & Incentives — AI models might change the incentives for creating new frameworks, libraries, and languages. In turn, this might change the open source model forever.
There is a lot to unpack. Let’s dive in!
Hey 👋 this is Luca! Welcome to a new 🔒 weekly edition 🔒 of Refactoring.
Every week I write advice on how to become a better engineering leader, backed by my own experience, research and case studies.
Here are the latest articles you may have missed:
To receive all the full articles and support Refactoring, consider subscribing 👇
You can also learn more about the benefits of a paid plan.