I mostly agree with your assessment of how OOP is viewed today. In many technical circles, inheritance is seen as actively harmful, and polymorphism is at best tolerated and often misunderstood. The hype is largely gone. I’ve also heard the same managerial rhetoric you mention, where OOP and UML are treated as unquestionable defaults rather than design choices, so that part unfortunately still resonates.
Where I disagree is on encapsulation being the “good” part of OOP.
Encapsulation, as a general idea, is positive. Controlling boundaries, hiding representation, and enforcing invariants are all valuable. But encapsulation as realized through objects is where the deeper problem lies. Objects themselves are not modular, and the act of encapsulating a concept into an object breaks modularity at the moment the boundary is drawn.
When you encapsulate something in OOP, you permanently bind state and the methods that mutate that state into a single unit. That decision fixes the system’s decomposition early and hardens it. Behavior can no longer move independently of data. Any method that mutates state is forever tied to that object, its invariants, and its lifecycle. Reuse and recomposition now operate at the object level rather than the behavior level, which is a much coarser and more rigid unit of change.
This is the core issue. Encapsulation in OOP doesn’t just hide implementation details; it collapses multiple axes of change into one. Data representation, behavior, and control flow are fused together. As requirements evolve, those axes almost never evolve in lockstep, but the object boundary forces them to.
What makes this especially insidious is that the failure mode is slow and subtle. OOP systems don’t usually fail immediately. They degrade over time. As new requirements arrive, developers feel increasing resistance when trying to adapt the existing design. Changes start cutting across object boundaries. Workarounds appear. Indirection layers accumulate. Eventually the system is labeled as having “too much tech debt” or being the result of “poor early design decisions.”
But this framing misses the point. The design mistakes were not merely human error; they were largely inevitable given the abstraction. The original object model could not have anticipated future requirements, and because it was not modular enough to allow the design itself to evolve, every change compounded rigidity. The problem wasn’t that the design was wrong. It’s that it was forced to be fixed.
Polymorphism doesn’t fundamentally resolve this, and often reinforces it. While polymorphism itself is not inherently object-oriented, in OOP it is typically expressed through stateful objects and virtual dispatch. That keeps behavior anchored to object identity and mutation rather than allowing it to be recomposed freely as requirements shift.
The deeper requirement is that a system must be modular enough not just to extend behavior, but to change its own design as understanding improves. Object-based encapsulation works directly against this. It locks in assumptions early and makes architectural change progressively more expensive. By the time the limitations are obvious, the system is already entangled.
So while I agree that inheritance deserves much of the criticism, I think encapsulation via objects is the more fundamental problem. It’s not that encapsulation is bad in principle. It’s that object-based encapsulation produces systems that appear well-structured early on, but inevitably accumulate rigidity and hidden coupling over time. What people often call “tech debt” in OOP systems is frequently just the unavoidable artifact of an abstraction that was never modular enough to begin with. OOP was the tech debt.
The way forward is actually simple. Avoid mutation as much as possible. Use static methods (aka functions) as much as possible. Segregate IO and mutation into its own module separate from all other logic. Those rules are less of a cathartic paradigm shift then OOP and it takes another leap to see why doing these actually resolves most of the issues with OOP.