The Infinite Software Crisis
Today I watched a talk by Jake Nation:
“The Infinite Software Crisis”.
It was a very balanced take, with a lot of good insights into the kind of future we might be heading into.
The point isn’t that everyone is going to lose their jobs to AI because it becomes some all-powerful superintelligence that makes us redundant (even though that might still happen at some point).
His take is more grounded. I’ll mix that into my own thoughts here — but if you want his exact perspective, watch the video.
Before, we did the entire loop of software development ourselves.
Everything lived in our heads, and we hand crafted it bare hands.
Now we have these “autocomplete machines” that can write code faster than we ever could.
That forces us to change our strategy. Not because we want, but because we have to.
Instead of jumping straight into the code and starting to write and debug, we need to step back and own the problem first.
We need a clear specification of what should be built — not line-by-line in code, but in words.
That means understanding:
- the problem itself
- the codebase
- the dependencies
- the requirements
- the domain we’re working in
From there, AI becomes a tool.
A means to an end.
This also means we still need to use our brains.
We still need to understand how code works.
We still need to think critically.
We still need to have strong opinions about structure and quality.
Even now, with agents getting faster and more capable, this is still very true.
If you’ve worked in a complex or legacy system and used AI tools, you’ve probably felt this already.
What the future brings, we don’t know.
But this is the mental model I’m using right now:
Own the problem.
Define it clearly.
Understand the codebase deeply.
Always increase your knowledge and understanding.
Lean into the new workflows and tools.
Let AI handle more of the execution.
I’ll ride that model for as long as it holds up.