The last time I wrote about the use of AI in software engineering, I was shitting on Devin.
It turns out, I was half right. Devin evolved a bit from what it originally was, bounced around industry news like a beach ball at a Bilmuri concert, and now you probably haven’t heard of them much since the Windsurf debacle. But, to their credit, they are still around. I still think Devin is weird.
Then something else happened shortly after that caught a lot of people’s attention in May and got even louder in November. Claude Code entered the chat.
The Great Awakening™
Opus 4.5 launched in November of 2025, and things started to get even more weird. The internet was abuzz with people saying it was possible to generate actual production-worthy code with frontier models.
Some people were elated with the possibilities of working with agents this way, and a lot of people kept on doing their thing like nothing had happened. I was in the former camp. Back in 2023, I quit my job to return to IC work because of my excitement around LLMs.
This was a big moment.
Okay cool, I guess this is real
Right after Opus 4.5 launched, all of my most AI-pilled friends were talking.
DUDEEEE
So I bumped up my Anthropic subscription, read some papers that had been recently released, and ran some experiments trying to see what the fuss was all about.
I was absolutely floored. Opus 4.5 and Claude Code were pretty damn good at tackling the same shapes of problems I did at my day job with a little bit of effort.
During this time, I wasn’t a stranger to working with agents to generate code. I paired with them all the time. Claude was my buddy, and it helped me navigate hard problems. But this was very different. Normally, I would work side by side with an agent and get its opinion on how I was approaching problems and the code I was writing. Sometimes it would generate something that was passable or had the spark of something I would actually write myself, and it was helpful. My hands were always the ones on the keyboard, though.
This was the first time it generated something that I would approve if I saw it in a pull request. It wasn’t perfect, but it was good.
The journey begins
Fast forward to now. I haven’t crafted a single line of artisanal code since May. In the lead-up to that, most of the time I did actually edit or write code, I did it just to feel something. I was in deep grief, and the irony of the situation was that I was working on systems that would make the very thing that was replacing a strong skillset of mine more effective.
I’ve been a professional software engineer for over half my life and enjoyed programming well before that. Hell, both of my parents were programmers. It will always be a huge part of my identity.
I am The Code Man.
The reality that a lot of us have had to come to terms with is that a system with impressive algorithms became better than us at doing one of the things we love most in this world. It’s a giant bummer.
Or is it?
If I had to guess, this is the question on the minds of a lot of people who aren’t in full-on AI Delirium right now. It’s definitely on mine.
Our industry is aflame with people talking about the shape of engineering now that AI has won. You can sense some sadness in the room. Things have to change. The old bottlenecks apply no longer, and quality code is cheap. People online give out YNGMI advice like their constraints are the same shape as everyone else’s building software. It’s a hell of a lot to navigate.
Where am I going with this?
I’ve had my ups and downs just like everyone else this past year. I’ve been through cycles of deep thought, trying to find joy in my work, and figuring out where I fit in the new landscape of the industry I love. The big things that have been rolling around my head, half-formed, are taking shape. I have a lot of words in me that might just help other people, and that excites me.
We are just getting to the point where loud voices in the room are discussing just how overindexed our industry was (is?) on AI replacing human labor. But the reality is that there is an incredible amount of value to be captured working with these tools when they are paired well with human judgement.
This is indeed the most exciting thing that will happen in our lifetimes for computing
And guess what, buddy? If you’re reading this and the things I’m saying resonate with you, you’re part of the story here. That’s objectively very cool and should excite you.
Loss is nothing else but change, and change is Nature’s delight.
The shape of your work is going to change. It probably already has for you. It could all come crashing down at a moment’s notice, but for now, it’s up to you to make the best of it and lead your teams through uncertain times if you want to be part of the story unfolding.
I think I might be able to help with that, so I’m writing this post to tee up a flurry of other posts that deserve their own space. I’m going to talk about:
- The grief you might be feeling right now, and how I made sense of it
- My personal ethical stance on AI at large
- The changes in how we build compelling, useful applications as a team
- How to think about your new place in the Software Development Lifecycle
- How to work with agents and make them not suck
You might be thinking I’m a little late to the party here. I think I’m right on time.
Stay tuned.