On "Quoting Ted Sanders"

Model size is doing most of the explanatory work for LLM capability, and that's both reassuring and dangerous for anyone betting on differentiation. It means you can stop obsessing over which model is marginally better at your specific use case—scale solves most problems—but it also means every vendor is chasing the same lever, so competitive moats erode fast. For practitioners, this should kill the urge to hand-tune prompts for Claude versus GPT versus whatever's next. Pick one, build on it, and expect it to become obsolete in months.