Eulogy for the Software Engineer

Summary

In this eulogy, software engineer Dave Kiss argues that AI-assisted development has changed the work of programming, but has not made the software engineer obsolete. What is ending, he says, is “the hard way being the only way”: manually writing every line, solving every problem alone, and relying on long stretches of uninterrupted coding.

Kiss recalls how the profession was built on open access and shared learning—from online forums and public documentation to open-source tools. He acknowledges the grief and uncertainty caused by AI systems completing in minutes tasks that once took hours or days. But he argues that coding was genuinely difficult, and that engineers’ judgment, persistence, and accumulated taste are not captured simply by collecting their public code and answers.

He describes three familiar tools—and what they represent:

• Headphones: The flow state of concentrating deeply on a problem.
• Notebook: The sketches and plans that define what a system should do.
• Keyboard: The act of personally typing the code that ships.

AI changes how each fits into the job. Engineers may not write or read every generated line, but they still need to understand the problem, define success, verify results, and take responsibility for the shipped work.

Practical guidance

Kiss describes merging about 3,000 generated lines into a command-line tool without reading every line. He says the important work had happened earlier: the team had defined what “done” meant and established a standard the changes had to meet. He then ran all 17 new commands against a live Mux account and checked their outputs before merging.

His proposed approach to AI-assisted development is to:

1. Understand the problem deeply enough to recognize a good solution.
2. Define “done” before building, rather than relying on vague expectations or superficial review.
3. Verify behavior against meaningful checks, such as running commands and evaluating their actual outputs.
4. Stand behind what ships under your name and be able to explain your process and why your checks were sufficient.
5. Document project context and constraints in the repository, such as its purpose, boundaries, rules, and rationale. Where a rule is critical, encode it deterministically so that an ignored instruction cannot easily lead to the wrong result.

He sees AI as an amplifier: strong teams can deliver much faster, while teams with poor habits can reach dead ends faster. The technology increases the importance of engineers’ taste—their perspective on what to build, what to reject, and what counts as good. Because that judgment is not automatically supplied by AI, experienced engineers should deliberately pass it on by teaching, sharing decisions, and mentoring. Newer engineers should ask questions and practice defining success for a task before asking AI for help.

The talk also stresses keeping the profession accessible. Kiss cites Aaron Swartz’s commitment to freely shared knowledge and warns that paid AI subscriptions could become barriers to entry. He hopes open models and free access will help ensure that cost does not determine who can become a software engineer.

Reviews, guides, or tutorials

No product review or step-by-step tutorial is presented. The talk offers practical guidance for working with AI-generated code and mentoring engineers.

Main speaker and source

Dave Kiss, speaking at a Resend event. Historical figures mentioned as influences include Grace Hopper, Margaret Hamilton, Tim Berners-Lee, Linus Torvalds, Joel Spolsky, Jeff Atwood, why the lucky stiff, and Aaron Swartz.