Data
AI
The Stack Underneath
Hello, we are BRILLIAN’S. Trying to make an effort to put the right people for you to get the best results. Just insight !
The common belief is that best practices are about quality. That belief survives because quality sounds vague. In reality, best practices benefits developers in the long, they are not only about aesthetics.
If you are optimizing for speed alone, in fact, many best practices may slow you down.
The advice to “keep code simple” is widely repeated and widely misunderstood.
Most teams think simplicity means fewer lines or clever reuse. That belief persists because it looks efficient in code reviews. It is also wrong.
Simplicity means fewer concepts that must be understood at the same time.
I have seen systems where a single change required opening twelve files, each individually “clean.” Nothing was broken in isolation. The system was broken as a whole.
What breaks when you ignore simplicity is not correctness. It is confidence. Engineers stop making changes without defensive layers of checks and reviews. Delivery slows even though nothing obvious is wrong.
When simplicity is enforced, changes feel local. Engineers move faster later, not sooner.
You have probably heard “don’t repeat yourself” treated as law. It is not. Some duplication is cheaper than the abstraction required to remove it.
The error teams make is not duplication itself. It is letting duplicated logic evolve independently without noticing.
I have watched teams fix the same bug three times in three services because “they are slightly different.” They were not different in the way that mattered. They were different because nobody owned the shared behavior.
If you get this wrong, bugs reappear and trust erodes. If you get it right, fixes stay fixed. That alone justifies the effort.
This is where strong engineers often overestimate themselves.
Dense code can be correct, fast, and well-tested. It can also be hostile to the next person who has to modify it. That person is often you, six months later, without the context you assume you will remember.
Clever code survives only when the author stays around. Teams do not.
When this goes wrong, parts of the codebase become untouchable. When it goes right, code changes hands without ceremony.
If you think this is about style, you have not maintained a system long enough.
Many teams pretend version control is neutral infrastructure. It is not. It encodes accountability.
When commits mix unrelated changes, you remove the ability to reason about cause and effect. Postmortems turn vague. Rollbacks become risky. People argue instead of diagnosing.
This belief that “commit history doesn’t matter” survives because teams only feel the pain during incidents. By then it is too late.
If this practice is wrong, you lose traceability. If it is right, failures become explainable. That difference matters under pressure.
Most explanations frame modular design as elegance or reuse. That framing misses the real benefit.
Modules exist to limit damage.
In tightly coupled systems, a small change forces a full retest. Engineers become cautious. Releases slow. Nobody can tell which parts are safe to touch.
In systems with real boundaries, mistakes stay local. That is the point.
If this practice fails, everything depends on everything else. If it works, teams can move independently without coordination rituals.
The belief that tests “ensure quality” refuses to die. It survives because it sounds reassuring.
Tests exist to preserve intent over time.
People forget why code behaves the way it does. Tests capture that knowledge imperfectly, but better than nothing. When behavior changes accidentally, tests surface it while context still exists.
What breaks without tests is not correctness on day one. It is correctness after the tenth change.
If you think tests slow you down, wait until you refactor without them.
Automating builds and releases too early is wasteful. Automating too late is painful. There is no universal threshold, and anyone who gives you one is guessing.
I have seen teams automate pipelines before they understood their own release process. They locked in confusion. I have also seen teams rely on hero engineers for deployment until those engineers burned out.
What is unknown here is timing. The right moment depends on release frequency and failure cost. My view would change if failures were cheap and rare. They usually are not.
Principles like single responsibility and dependency inversion are useful under pressure. Applied early, they create indirection without payoff.
The belief that “clean architecture from day one” is responsible survives because it signals seriousness. It also slows learning when requirements are unstable.
If you get this wrong, you ship less while feeling disciplined. If you get it right, structure appears only when it earns its cost.
Many readers believe best practices are about professionalism. That belief persists because it flatters teams that follow them.
They are not about professionalism. They are about survival under growth.
Teams that ignore this frame best practices as optional polish. Teams that learn it too late discover that rewrites are also a practice. An expensive one.
Best practices do not make software good. They make bad outcomes less likely.
If that sounds unambitious, it is. Software fails in boring ways. The job is to make those failures rarer and cheaper.
If this view is wrong, the cost is extra ceremony. If it is right, the benefit is years of avoided pain. I am comfortable with that trade.
You might also like
)
