Musings on the Evolution of Software Engineering in the AI Era

By Andrew Diamond
Published: June 28, 2026

To provide some context for those unfamiliar with my background: when I am not immersed in the world of novel writing, my professional life is dedicated to software engineering—essentially, I spend my days writing code.

Currently, the software industry is deeply intertwined with artificial intelligence. AI has become exceptionally proficient at programming for several specific reasons:

---

The Paradigm Shift in Workflow

Before the integration of AI, the development cycle was a rigorous, manual process.

The Legacy Workflow:

1. Requirement Gathering: Receive a request for a new feature.
2. Specification: Draft a formal document detailing the feature's scope, user access, and success criteria.
3. Architectural Research: Investigate the ideal $\text{algorithms} + \text{data structures} + \text{libraries}$ to implement the solution.
4. Implementation: Write the actual source code, create unit tests, and author documentation for both users and future maintainers.
5. Peer Review: Submit a pull request for colleagues to critique and approve.

Now, the process has been streamlined into a more supervisory role.

In this modern approach, you provide a prompt to an AI agent, review the generated output, and make adjustments—either manually or by asking the AI to iterate.

The Core Change: In the previous era, the creative heavy lifting occurred within the human mind. Today, the developer supervises a creative process occurring within the "black box" of the AI's neural networks.

While you still spend effort crafting a precise and accurate prompt, the "hard thinking" associated with manual coding is largely absent. You have transitioned from a creator to an editor.

The Necessity of the Senior Engineer

AI can generate syntactically correct code, but it lacks a holistic understanding of the project. It is blind to the "big picture." Specifically, AI is unaware of:

• Legal Constraints: Whether the code violates specific industry regulations.
• Performance Nuances: If a request will take $10\text{ms}$ or $10\text{minutes}$ to execute.
• Future Roadmaps: How this code might clash with a feature a teammate is building three weeks from now.
• Security Interdependencies: If a new function creates a vulnerability when paired with a sensitive function written months ago.

A senior developer possesses this institutional and systems-level wisdom. To them, an AI is like a highly efficient junior or mid-level developer: fast and generally competent, but lacking the twenty years of deep experience required to ensure the system remains stable and secure.

---

The Parallel: Software Engineering vs. Novel Writing

To illustrate this, consider the process of a historical novelist. Imagine a scene where two political figures are debating outside St. Paul's Cathedral. The writer doesn't just start typing; they perform deep research:

RESEARCH NOTES:
- Attire: What would a 19th-century diplomat wear vs. a merchant?
- Atmosphere: What did the London fog smell like in 1840?
- Politics: Who are the current ministers? What is the tension regarding the Corn Laws?
- Recent Events: Which treaty was signed last Tuesday that fuels this argument?

The novelist then weaves these factual anchors into a fictional narrative. This is remarkably similar to software engineering; a seasoned developer doesn't guess—they rely on years of knowledge regarding caches, databases, and algorithms before they ever touch the keyboard.

The State of "Flow"

Both the coder and the novelist experience a profound psychological immersion. This is what psychologist Mihaly Csikszentmihalyi termed "Flow"—a state of optimal experience.

In this state, time becomes elastic. The mathematical perception of time shifts: $\Delta t_{perceived} \ll \Delta t_{actual}$

It is common for a developer to check the clock, dive into a complex problem, and realize that while they felt only ten minutes had passed, four hours have actually vanished. This is why many engineers prefer working late at night; the absence of corporate interruptions allows them to reach this peak creative state.

A Final Thought Experiment

Imagine the historical novelist is suddenly offered a deal by her publisher. They want to increase her output from one book every two years to four books per year.

To achieve this, the publisher hires a fleet of inexpensive, competent college students to write five pages a day. The publisher still wants the books to maintain the novelist's signature quality, so they keep her on—not as the primary writer, but as the editor.