Three things changing at three speeds: software development (fastest), people (middle), organizations (slowest). The gaps are growing.
The natural response is alarm. I think that makes it worse - people retreat to the familiar harder.
AI has to be told what to do. The human thinking about what to build matters more now, not less. The teams adapting fastest are the ones where it is safe to not yet know.

The Bottleneck Is Not the Tool
https://einbacker.de/essays/the-bottleneck-is-not-the-tool.html

There's a quiet risk with experience: becoming rigid.

With agentic software development, I've noticed experience doesn't automatically help—but when it stays adaptable, it compounds.

What matters now:
• delegation instead of control
• judgment over throughput
• experimentation without attachment
• writing clear intent
• staying curious even when you "should already know"

Wrote a reflection on what still needs practice.

https://www.einbacker.com/essays/experience-compounds.html

I've found AI helps most before you know what to build—when the problem is still fuzzy.

Using agentic tools less as "answer machines" and more as thinking partners:
• exploring tradeoffs
• surfacing risks
• simulating decisions before committing

What hasn't changed: ownership. Someone still decides what matters, what to keep, and what to throw away.

https://einbacker.com/essays/thinking-with-ai.html