Ofline
Writing code is easily done, but identifying the code that ought to be available is what's difficult nowadays.
Many experienced engineers are having a conversation regarding the capability of "taste" to be the most crucial ability of the AI age. "Taste" here doesn't refer to fashion. Instead, it’s about judgment. It is about being able to assess the AI output and understand precisely what’s inappropriate about it, the type of thing that will add up as technical debt, before it's pushed through.
Why This Matters Right Now
Junior-level AI virtually generates code for free thanks to AI. You can develop a feature immediately that would have needed an entire day to build. But being "functional" and being "ready to ship" are not equivalent to each other.
The space amid those two terms is where skilled engineers show their real value. Or where they reveal they have been slacking off.
Taste Is Not Vibes
I am referring to taste, which is how to distinguish code that needs to be written. 2 decades of experience, and here's what I will reveal about this taste concept. It's not some type of cryptic, elusive instinct. It's a pattern recognition that develops as you observe programs succeed and fail for years.
→ Knowing that a 200-line abstraction will save you now but cost you in six months.
→ Recognizing when AI output follows the "happy path" but ignores every edge case your users will hit.
→ Feeling the friction in an API surface before anyone files a bug.
There is an ongoing debate: does having good taste is enough or it should be combined with solid architecture knowledge? In my opinion, having good taste without architecture knowledge is mere personal preference. Having architecture knowledge without good taste leads to unnecessary complexity. You must have both. However, there is a lack of good taste right now because no one is educating for it.
The Role Shift Nobody Prepared For
Earlier, senior developers were appreciated because they could build things quickly and well. Now AI builds things quickly. The "well" part depends entirely on you.
Your role is more similar to a film director than a construction worker. You're inspecting, selecting, altering. You will have to reject 80% of automatically generated content and understand the reason behind it. This skill is completely different from writing code independently.
If your only advantage is typing speed and syntax, you are already in distress. In fact, these were never senior skills — it's just that the industry has allowed some people to think that way. 🫠
How You Build Taste
You can't read a book about it. You build taste by shipping things, watching them break, and developing strong opinions about why.
→ Review more code than you write. Especially code that's been in production for a year.
→ Study systems that aged well. Ask what decisions made them resilient.
→ When AI gives you output, don't accept it — interrogate it. What would you change? Why?
The engineers I respect most can look at a PR and say "this works but it's wrong" and then articulate the reason in one sentence. That's taste. It's specific, defensible, and earned. 🎯
The Uncomfortable Truth
Some senior engineers are going to struggle here. Not because they lack knowledge, but because they never developed the editorial instinct. They were builders, not curators. And the industry rewarded building for two decades.
That era is ending. The 10x engineer of 2025 isn't writing 10x more code. They're preventing 10x more bad code from shipping.
So here's what I want to know: Do you think "taste" is a real, trainable skill — or is it just gatekeeping dressed up in a new outfit?