My AI journey so far

Like most developers I think I’m finding my way with AI.

On a day to day basis, nothing feels that out of the ordinary, but when I look back over the last year, my role has changed more than at any other point in my career.

I’ve been doing this stuff for 30 years, so I’ve seen a fair amount of change: CDs/DVDs, internet, git, cloud, CLOUD, SAAS, FAAS, PAAS, social media, microservices, Kubernetes. But even the switch to cloud - which was pretty seismic - was not like this.

A year ago, I was copying and pasting things between an AI chat and my IDE. I tried vibe coding before it was a thing and concluded that it didn’t work (it didn’t at the time). I used it for auto completion, which was pretty cool.

Then around the start of the year a developer colleague mentioned to me that they hadn’t written a line of code for 2 weeks. I couldn’t understand what they were saying. What do you mean “you haven’t written code in two weeks?”. And then, just like that, 2 weeks later, I was in the same place as him. I also was now not writing code.

Since some point earlier this year, I’ve barely written any code. All of my time has been taken up with prompting, learning how to prompt better, trying out different agents and skills, writing custom agents and skills, working out how to evaluate agents and skills, reviewing code, talking to colleagues about all of these things and - at the same time - shipping (hopefully) working code.

The amazing thing is, for the most part, the code works. The even more amazing thing is that, for the most part, the code is good!

I’ve read a lot online about how the code that AI produces is rubbish and I can only conclude from this that I have very low standards.

There is a huge amount of churn at the moment, a huge amount of change, of uncertainty. Things are changing so fast. Many things are changing at once, making measuring stuff really hard. It feels like riding a wave (I presume so anyway, I’ve not actually ridden a wave), or riding a sandworm, if Dune is your thing (again, presumption; I’ve not ridden a sandworm).

The current state of development is not for the faint hearted. I think it is a lot harder for juniors - they don’t have the scar tissue that experience brings and helps you keep agents on the right path.

I’m concerned about the pipeline of talent - companies seem more inclined to hire a few seniors who can work with AI, rather than invest in juniors. If that carries on, at some point the market will have to correct.

Personally, I’ve found this AI revolution a huge amount of fun. It’s made so many more things possible. I can deliver stuff that previously I didn’t have time to do or didn’t have the localised knowledge to deliver. Now I can bring my broader knowledge and let AI handle the localisation issues. I can also work on side projects that I would never have otherwise started - let’s build an Operating System - for fun! Why not!?

I know that there are broader challenges with AI - the effect on jobs, the reliance on a few Big Tech companies with questionable ethics, the effect of AI on the environment. But, purely from a developer perspective - it’s been like spending all afternoon riding the Big Dipper.

I don’t know what the next 12, 6 or even 3 months is going to look like; but I am excited to see how we can use these tools to design and build better software.