Does Code Quality Still Matter? It Depends on This — Max Kanat-Alexander, Capital One

Highlights

• All software success is determined by two factors: understanding the system and executing on that understanding.

• Modifications to a software system carry risks of introducing problems and making future changes harder, both of which stem from difficulties in understanding.

• Disposable software represents a new category where consequences of failure are minimal, replacement cost is near zero, and maintenance cost is negligible, rendering traditional code quality practices unnecessary.

• Durable software systems will continue to be critical and their absolute volume will increase because AI coding tools make creating them easier.

• Standardization, in-workflow tooling, automated testing, and maximum linter strictness decrease the overall cost of engineering rigor.

Timeline

1. 00:00 - 04:41

   The Core Purpose of Code and Software Understanding

   Software engineers frequently rely on the noncommittal answer that everything depends on context. To provide a rigorous basis for reasoning about software, all goals such as scaling systems or aligning models trace back strictly to comprehension and execution. While complex codebases like Chromium possess infinite potential system-wide behaviors, structured components allow engineers to comprehend specific sections like URL parsers.

   • Every objective in software development relies entirely on understanding a system and executing on that understanding.
   • The primary purpose of writing code is to establish understandable structure rather than just functioning as a raw series of machine instructions.
   • Engineers must grasp two primary concepts: the actual behavior of the system and how to modify that behavior safely.
2. 04:41 - 10:50

   Risks of Modification and the Rise of Disposable Software

   Every system modification risks budget overruns, unmet user requirements, or elevated future risk, which stem from underlying comprehension gaps. Systems eventually reach a tipping point where every modification creates cascading problems. However, AI tools enable a massive surge in disposable software where traditional engineering rigor does not apply because components can be replaced in hours or built from scratch for single-use tasks.

   • Modifying code carries risks of introducing immediate problems and increasing the difficulty of making future changes.
   • The difficulty of understanding software compounds in a non-linear fashion over time as modifications accumulate.
   • Disposable software emerges as a new category where failure consequences, replacement costs, dependencies, and maintenance requirements are minimal.
3. 10:50 - 16:15

   Reducing the Cost of Engineering Rigor

   While durable software will remain critical and grow in absolute volume due to AI acceleration, teams can manage the spectrum by lowering the friction of applying engineering standards. Standardization removes unnecessary options like multiple programming languages or frameworks, while automated in-workflow tools execute quietly without manual intervention. Ultimately, code quality matters for durable software and does not matter for disposable software, but the cost of achieving quality can be systematically reduced.

   • Disposable and durable software exist on a spectrum defined by consequences, replacement cost, dependencies, and maintenance cost.
   • Decreasing the cost of engineering rigor allows developers to easily transition systems from disposable to durable when requirements change.
   • Standardization, automated in-workflow tooling, simplified testing infrastructures, and maximum linter strictness eliminate redundant decisions and reduce the expense of rigor.