The conclusion to this essay doesn't make a convincing case. How do we know that academics are doing things correct when "sparks of AGI" is two years old and we don't have serious agentic software? How can you assert that isolated knowledge improves AI cognition when LLM performance directly correlates with the volume of training data? How could anyone actually prove that "academic silos" are challenging old paradigms when we aren't allowed to see what's inside them? What heuristics are you using? Why can't we see them?