You restate that disruptive innovation happens when gains are large enough to overcome inertia, and that smaller conceptual shifts aren't worth pursuing. Your premise is pragmatic: if it mattered, the market would already have adopted it.

This still sidesteps the article's point that what we measure as "efficiency" is itself historically contingent. GPUs succeeded precisely because they exploited massive parallelism within an already compatible model, not because the ecosystem suddenly became open to new paradigms. Your example actually supports the article's argument about selective reinforcement.

There's nothing to agree to disagree about. You're arguing a point the article does not make.