AMI Labs explicitly pursues non-LLM, JEPA-style world models 2 sources. Because this structure blocks direct competition on standard text leaderboards, resolution requires independent expert validation. Aligning with peers taking paradigm-shifting bets and related questions modeling a delayed commercialization path, there is a heavy concentration of outcomes in the 2030s, as novel non-LLM architectures will require significant time to match incumbent frontier generality. The 60% probability of incumbent dominance implies that matching the generality of OpenAI or Anthropic models with a novel architecture will take over a decade, placing the median at June 30, 2038. Even a 25% rapid-breakthrough scenario leaves early percentiles from June 30, 2029, to December 31, 2032, due to the baseline difficulty of the task. A 15% probability of a permanent pivot to specialized layers or an outright failure to generalize maintains a heavily extended right tail reaching January 1, 2049, for the 75th percentile, and potentially never occurring for the 90th percentile.