An article in EnvironmentalResearchWeb oversimplifies the relationship between simple and complex climate models.
For one thing, it is not news that reduced complexity models that run on desktop computers can be built, especially at the global average scale. (More typically for research, they are averaged by latitude.)
For another, without the complex model, it would be difficult to build the reduced complexity model, as the observational record is limited. As time goes on and the number of observations increases, this is somewhat mitigated, but it’s still easier and far less issue-laden to tune your simple model to a complex model where you have unlimited perfect quasi-observations, rather than real world, noisy measurements subject to drift.