Code is Cheap. Your Senior Developers Aren't.

I regularly speak with founders and CTOs who say their teams aren't coding with AI yet. Too risky. Too unpredictable. I also hear developers say they're proud to never use it because the quality is too low.

At the same time, those same senior developers are rolling out internal dashboards, CSV importers, form flows, and standard API integrations.

That's not caution. It's a sign that your most expensive talent is doing the work with the lowest leverage.

Code is Now Cheap, Software Isn't

Code used to be expensive. That's why companies hired so many engineers. That era is over.

Tools like Claude Code, Cursor, GitHub Copilot, and OpenClaw (the newest sensation) make typing cheap. Software remains expensive because the real costs lie elsewhere:

• understanding the problem
• making architectural choices
• integrating with existing systems
• handling exceptions and failure scenarios
• taking ownership of maintenance and long-term impact

Boris Cherny shared that a recent month of his work on Claude Code was written by Claude Code and Opus 4.5, with hundreds of pull requests and tens of thousands of changed lines of code. That's about throughput, not quality. Output is no longer the bottleneck.

The bottleneck is still building solutions that connect to real problems, not just shipping code.

Code is cheap. Software isn't.

Andrej Karpathy describes vibe coding as a shift where you can spin up software for a single task and throw it away. That raises the premium on framing the problem and evaluating it; exactly where senior engineers add value.

The Evolution of Abstraction

Every generation of programming languages raised the abstraction level: from assembly to high-level languages to managed runtimes. AI is the next step. We specify intent and context; the system fills in the 'how'.

Typing was never the job. Thinking was.

Typing Work vs. Thinking Work

Peter Steinberger writes about shipping at inference speed. Output is now limited by inference time and deep thinking, not typing. Most apps still move data from one form to another. The leverage is in deciding what matters.

My view: AI collapses implementation costs but raises the bar for judgment. That makes experienced engineers more valuable, just in a different way.

You don't pay senior developers to type. You pay them for their judgment.

The New Role of Your Dev Team

This isn't a call to replace developers. It's a call to deploy them differently:

• from writing code to system design
• from implementation to orchestration, with AI as a faster junior
• from building to deciding what to accept, reject, or revise

Karpathy notes that the profession is being refactored, with a new layer around agents, prompts, tools, workflows, and orchestration. The work shifts 'up stack', it doesn't disappear.

What This Means for Developers

Your impact no longer comes from lines of code. It comes from clarity, systems thinking, and the quality of your decisions.

You're also the ones who can quickly turn AI into internal solutions, which is now a competitive advantage.

What This Means for Founders and CTOs

If your team is still manually doing standard work in 2026, that's a management problem, not a technology problem.

One more implication: you can't reliably detect whether code was written by AI. Karpathy sees this in education: detection fails, so evaluation shifts to controlled environments. The same principle applies in software: assume AI-generated code exists. Don't invest in detection, invest in reviews, tests, observability, and ownership.

What Now?

1. Identify typing work: repetitive tasks that don't require unique expertise.
2. Introduce AI tooling as a multiplier, not a replacement.
3. Adapt the workflow to your team; there's no single right setup, use what fits your stack and constraints.
4. Reallocate time to architecture, integration, and difficult decisions.
5. Measure differently: not in lines of code, but in problems solved.

Teams that make this shift early will see their lead grow exponentially.

In Conclusion

Code is cheap. Specification is expensive. Judgment is scarce.

Your senior developers are too expensive for typing work. Let them focus on what needs to be built and why. The AI can do the typing.

Teams that ignore this will end up paying senior rates for work an agent does in minutes. Teams that embrace this build faster and better.

---

Interesting Reads

• Chris Gregori: Code is Cheap Now. Software Isn't.
• Peter Steinberger: Shipping at Inference Speed