A heavily discussed article on Lobsters makes a judgment, and we concur: AI programming tools accelerate code output, but the decision quality of system architecture is declining—because people are offloading problems requiring intuitive judgment to tools.
What this is
The "architect instinct" discussed in the article refers to the ability of experienced technical architects to quickly identify critical constraints, anticipate the consequences of tech stack choices, and judge which modules deserve investment when facing complex systems. This instinct isn't mysticism; it's pattern recognition accumulated through extensive practice.
The author's core argument: current AI-assisted programming tools (like Copilot, Cursor) excel at local code generation, but the core challenges of system architecture—how to slice module boundaries, how to design data flows, what technical debt is acceptable—still rely on human judgment. More dangerously, AI makes "writing code" too easy, encouraging a "code first, think later" tendency. Architectural issues get buried under massive amounts of quick-generated code, only surfacing when system complexity spirals out of control.
Industry view
Supporters believe the article pinpoints a key trend: AI lowers the barrier to coding, but raises the barrier to systems thinking. There are more people who can write code, but fewer who can make architectural decisions. One commenter compared it to how "after calculators became ubiquitous, mathematical modeling skills became even more important."
However, there are dissenting voices. Some developers argue that "architect instinct" is over-romanticized—much of this so-called intuition can actually be replaced by methodology and documentation. Architectural work in the AI era is indeed changing, but not necessarily regressing: AI-assisted prototype validation and architecture simulation are making architectural decisions more data-supported, rather than purely relying on experience-based "gut feelings." A sharper critique is that overemphasizing instinct may become a moat for veterans, detrimental to transparent team decision-making and the growth of newcomers.
Impact on regular people
For enterprise IT: When selecting technologies, "who writes code fastest" is no longer the core competitive advantage; "who can judge where system boundaries lie" is. The focus of hiring and team building needs to shift from coding speed to systems thinking.
For individual careers: The bargaining power of engineers who only write code is declining, but those who can understand business logic and make sound abstractions—whether they call themselves architects or not—are becoming more valuable.
For the consumer market: Ordinary users won't directly perceive architectural quality, but they will feel it indirectly—products with messy architectures iterate slower, accumulate more bugs, deliver worse experiences, and ultimately get eliminated by the market.