KnowledgeForge could route a reasoning task to the right mode. What it couldn't do was tell me if the output worked for a specific reader. That gap showed up the first time I tried to use it to evaluate outreach copy. I had a cold email. KF could