There's a phenomenon that's particularly intriguing—when evaluating infrastructure projects, people usually only look at the present. "Do we have this need now?" has become the biggest question. But often, the most impressive designs are precisely those that have already paved the way before you even shout "We must have this."

Take storage as an example. At first glance, it seems ordinary. "It's just storing data, what's so special about that?" That sounds reasonable, but the problem is, after an application runs for a year or so, what really starts to challenge you isn't speed, but the pile of historical data.

You begin to hesitate, afraid to change core logic for fear of breaking the existing trust system. Any modification risks causing a failure. At this point, you realize—whether the design is good or not is reflected in whether it can gracefully handle this kind of "immutable" historical accumulation.

Take Walrus's object model, for example. Essentially, it's designed in advance to defend against such scenarios. Objects aren't replaced; they evolve in place. History isn't overwritten; it continues to accumulate. A medium-sized application updating its state 3-6 times a day can reach 30-40GB in a year. You can't delete this data, and you must be willing to use it.

I believe that things like Walrus are not just some supplementary nutrients. On the contrary, they belong to those protocols that are difficult to adopt from a low level once you truly realize their value.