I think we're repeating the same mistake we made before operating systems existed.
I've been thinking a lot about the history of computing.
Before operating systems became standard, every application had to deal directly with hardware.
Memory management.
Device communication.
Storage.
Scheduling.
Every team solved the same infrastructure problems again and again.
Operating systems changed that.
They didn't replace applications.
They became the foundation that applications could rely on.
When I look at AI software development today, I see something strangely familiar.
Every company is creating its own:
- Prompt library
- AI workflow
- Coding rules
- Knowledge base
- Context management
- Agent orchestration
- Documentation structure
- Memory system
Most of this work isn't product differentiation.
It's infrastructure.
And almost every team is rebuilding it from scratch.
That makes me wonder...
What if we're still in the "pre-operating system" era of AI engineering?
What if the next big opportunity isn't another AI coding tool...
...but the common engineering foundation that all AI agents can build on?
That's the question that led me to start building NAEOS.
I don't know whether this idea is right.
But I think it's a question worth exploring.
History has shown that standards often create more innovation than individual tools.
Maybe AI engineering is approaching a similar moment.
I'd love to hear other perspectives.
Do you think AI development will eventually converge on shared engineering standards, or will every company continue building its own stack?
I like the historical analogy because it shifts the question from "What should AI do next?" to "What should AI stop rebuilding over and over?"
One thing that came to mind is that a technology ecosystem starts to mature when more teams agree on what shouldn't be a competitive advantage anymore. Once a problem becomes shared infrastructure instead of something every team reinvents, it frees people to compete on the things that actually make their products different.
Whether AI ends up with one foundation or several, that transition feels like an important milestone.
That's a great way to frame it. The real inflection point isn't necessarily when everyone agrees on a single standard, but when certain capabilities become "table stakes" rather than strategic differentiators. At that point, organizations can stop spending energy rebuilding the same foundations and redirect it toward domain expertise, product design, and customer value. That's usually a sign an ecosystem is starting to mature.