The Non-Deterministic Mindset Shift: From 30 Lines of Code to Orchestrated Natural Language

How to operate in a non‑deterministic world: move from unit-level certainty to system-level observability, evals, and human-in-the-loop checks when orchestrating natural‑language driven components.

Sep 29, 2025 0m 22s

James LePage

Expert Insights

James LePage contends that AI-enabled software development necessitates a shift from deterministic to non-deterministic thinking. Incorporating language processing elements, traditional systematic testing methods may no longer suffice.

Hear James explain:

The transition from understanding straightforward code lines to managing layered natural language systems.
How this liaison of code and natural language contributes to a larger system that orchestrates structured outputs.
The potential challenges and transformation to the testing process, which now has to deal with the complexity of natural language processing.

We have engineers going from, I have written 30 lines of code, and now I have this new feature and I know exactly how it works, and I know exactly how to test it too. I have this 30 lines of code wrapping, a bunch of natural language, contributing to a more global system that orchestrates this natural language into trying to create a structured output that I can then apply back to my product.

— James LePage, Director of Engineering AI, Automattic

THE NEW DEFAULT angle

Here's how software teams can upskill and adapt processes for effective AI-assisted development:

Foster Cross-Competency: Encourage your software engineers to grasp not only coding but also the intricacies of natural language processing, making them prepared for the combination that AI development entails.
Redefine Testing Approaches: Traditional testing methods may lack efficacy with AI systems that involve complex natural language handling. Teams should adapt their testing strategies to assure quality in such non-deterministic systems.
Build Systemic Understanding: Developers now work with code that feeds into a larger, interwoven system. It's essential to cultivate an understanding of this broader architecture and how individual code contributions fit into it.
Embrace Learning & Iteration: The shift to AI-assisted software development may pose challenges initially. Proactively nurturing a learning mindset and embracing iterative improvement can help teams navigate these complexities.