I’m changing Wilder Code
Being kind to myself, and adapting to my own changing world.
It’s been a couple of weeks since I last posted (I was galavanting around the Scottish Highlands for 2 weeks), and this is a different type of post, but I wanted to formalise my plans to change some of the content I write about here.
I still want to write about scalable code, but there’s a lot more I’d like to write about and share.
Why change?
I’ve seen a shift in my perception to writing “scalable code” over the last few months. And yes, it’s due to AI.
I’ve found myself writing less and less code recently. I still ship thousands of lines of code, but the amount that I write by hand is greatly diminished.
My role has shifted to reviewing code output by a LLM, and I’ve been adjusting how I work to optimise for review and trying to get the AI to achieve its goals quicker. And I’ve had great success with this.
But this shift in way I work has raised some questions as to how we get AI to generate what we want, and how we get comfortable with what it generates.
Where there’s a gap with AI
A little background to where my mind is at with some of these changes.
There’s many reasons why people think AI isn’t living up to the hype, but I feel one reason for this is our expectations when it comes to the quality of code we write.
It’s natural for us to feel the code we write is of the highest quality. We hold ourselves to a much higher standard than we hold others, and we care more about the code we write.
But we should not expect the same bar from our team mates. This isn’t a bad thing - we’re all the product of our own experiences and so we should expect others to write code differently to us.
It doesn’t mean the quality of what they’re writing is objectively worse, just it might not meet our own personal threshold for quality - and that’s okay.
I’ve previous written about the downsides of holding others to our own standards here 👇
But when it comes to AI, I think many of us hold it to the same standard as ourselves. Probably because the commits will have our name against it, and it presents itself as a representation of our capabilities as engineers.
But we shouldn’t expect AI to write code to our standards! It’s a product of it’s own experiences, and so we should look to do one (or both) of two things;
Treat AI as if it were another team member, and “lower” our expectations
Try to increase the quality of the AI output
So what’s changing?
Given the context above, I want to change some of my future posts to focus more on those two areas - both of these close that gap of pain I highlighted above:
How we can adjust our mindset and get comfortable with the “lower quality” output of AI (or even those around us)
How we can guide AI to increase the quality of it’s output
But there are other things that I’d just like to post about that I feel others might find interesting. From deploying AI locally and saving a bunch of money on how it works, to how that could help your team/business, to side projects that solve interesting challenges and the decision process that led to their creation.
But to cover these to the quality standards I hold of my own posts (how topical), I need to be kind to myself and drop the weekly schedule for posting.
To summarise
I’ll be dropping the weekly schedule in favour of higher quality posts. I still want to post at least once per month, perhaps you’ll see weekly posts for a while, but for now, the schedule is relaxed.
I’ll be writing future posts to cater to AI more. Perhaps it will feature a section on “how to apply this for yourself” and “how to apply this for AI”
I’ll be branching out from “just” scalable code, writing about other things I find interesting that I feel my current audience will also find interesting.
I hope you will stick around, I think my best work is still to come 🙂