What sorts of apps are those? I tried testing various models with a test app as a benchmark, a local first app with CRDTs, and many, even frontier models, struggle heavily.