Agentic Engineering and the Sin of Incuriosity

You didn't get into this business to be indifferent to technological possibility

I’ve been trying to understand something about some of my fellow engineers. From where I stand, large language models and agentic coding tools represent the largest change in software in over two decades, and it’s not even close. As such, I figure that every software engineer in the business would, at an absolute minimum, have opinions about them. I’ve talked many times this last year with some extremely senior and very smart engineers, both in person and online. Some were wildly enthusiastic about agentic coding, while others were deeply skeptical. What I didn’t expect, and frankly am unable to understand, were the ones that were simply uninterested.

While I’m fully on Team Enthusiastic for agentic coding tools, I have no beef with Team Skeptical. Any engineer worth their pay has seen many greatly-hyped technologies turn to dross. Any engineer who has sat in on a vendor meeting knows that new technologies come surrounded by salesmen telling breathless bald-faced lies (or, if one is being charitable, “truths from the future”). Long-tenured engineers will have ridden the hype cycle of new technologies dozens of times, and seen just about all of the ways in which they can fail, some whimsical, some horrifying, most just kind of stupid and sad. After my own very long career, I’ve lived all of these more times than any three of my readers put together, and have absolutely earned my Olympic-class cynicism. Skepticism about a new technology is absolutely understandable, and indeed is a good part of what they pay engineers for.

To be absolutely fair, agentic coding tools present even more opportunities for skepticism than most new technologies. They have multiple hilarious failure modes, and those failure modes aren’t even slightly difficult to find. If you expected perfection from these strange new angels, you would quickly find yourself learning better. But requiring perfection isn't skepticism. It's an excuse, or possibly an indictment. We don't refuse to use databases because transactions can deadlock. We don't abandon Kubernetes because pods crash. We as engineers build out of components with imperfections all the time. There are entire bodies of knowledge like information theory and queuing theory that describe in precise mathematical terms just to what extent one can turn the fallible into the practical. We take brittle parts and build resilient wholes. At some point that's just what engineering is.

Yes, agentic coding tools need to be carefully prompted, but so do junior engineers. Yes, agentic coding tools sometimes hallucinate or get stuck in endless loops, but frankly so do I. Yes, the answers that come out of a large language model need to be carefully checked. But answer checking turns out be much easier to automate than answer generation. The most important mathematical question in computer science, P vs. NP, is basically referencing the fact that for a large number of questions, checking answers that you have been given is easy, even if generating the answers was irreducibly difficult.

I understand the skepticism, truly. I just don’t understand letting it stop you.

There’s a poem I keep coming back to by Kipling (a complicated and controversial figure for good reasons, and I’m not here to relitigate his legacy). He spent years around engineers building impossible things in impossible conditions, and ‘The Hymn of Breaking Strain’ captures something I’ve never found expressed better:

The careful text-books measure
(Let all who build beware!)
The load, the shock, the pressure
Material can bear.
So, when the buckled girder
Lets down the grinding span,
The blame of loss, or murder,
Is laid upon the man.
Not on the Stuff—the Man!

That’s the job. Knowing that materials fail, that systems fail, that we fail - and building anyway. This poem belongs to engineers now.

Incuriosity is the opposite of that. It’s not caution. It’s not wisdom earned from hype cycles past. It’s the decision to stop looking, stop wondering, stop building. And that, to me, is the sin - not skepticism, but the repudiation of what brought us here in the first place.

We are being offered tools that can regularly produce small miracles, somewhat tarnished, at the cost of pennies. If that doesn’t excite you, then I have to wonder what you came to engineering for. If your response to machines that produce wonders nineteen times out of twenty is to yawn, roll over, and go back to sleep until perfection is achieved, perhaps a new career is in order. The money men are paying you partially for your skepticism, but in the end they are mostly paying you to show them what technology can accomplish. If you’re not interested in that, it may be time to reassess your life choices.

In the end, if you are suffering the sin of incuriosity about agentic coding tools, I challenge you. Spend a weekend working with them. Don’t spend it looking for flaws, you’ll find those easily enough. Instead, spend that time looking for potential. Not looking for reasons to set the tools down, but looking for reasons to keep going. Don't stop until you have an idea of something that you could build with these tools that you couldn't build without them. Then build, and build some more. Keep building until you’ve astonished yourself.

---

This post was constructed as always with the able assistance of Claude Opus 4.5, who has the unique perspective of being both builder and built.