Source note
What Do Engineers Mean When We Say "Taste"?
Summary
The post argues that “taste” in software engineering is not a single mystic trait but an integrated calibration skill. It matters more as AI takes over mechanical coding work and leaves engineers with the judgment calls.
Problem
- Engineers use “taste” to describe the judgment behind good software choices, but the term is vague and has no clear decomposition.
- Current engineering education and literature cover coding, architecture, and testing better than product judgment and cross-cutting design judgment.
- As AI handles more routine implementation, the field needs a clearer account of what human engineers still contribute.
Approach
- The author decomposes taste into three linked parts: product thinking, system thinking, and quality-as-calibration.
- Product thinking means choosing what to build so it fits user needs, market needs, and technical reality.
- System thinking means choosing architectures that can survive change, failure, and future constraints.
- Quality-as-calibration means choosing where to be rigorous and where to move fast, based on context and blast radius.
- All three depend on knowing what is physically and technically possible, then calibrating across many tradeoffs at once.
Results
- No quantitative experiments or benchmark results are reported in the excerpt.
- The main claim is conceptual: taste can be broken into separable parts instead of treated as a single opaque trait.
- The post claims the strongest engineers coordinate product, system, and quality judgments at the same time, rather than following checklists.
- It also claims apprenticeship is still the main way taste transfers, because people learn the shape of the judgment from exposure to experienced engineers.
- The author argues that if AI absorbs most mechanical engineering work, taste becomes the main differentiator in software engineering.