Open-source models will likely close the performance gap with proprietary systems by mid-2026. Nathan Lambert argues that synthetic data and architectural refinements drive this convergence. While closed models maintain a temporary lead, the trend favors accessibility. Researchers should prepare for a shift where high-tier capabilities become commodity software rather than guarded secrets.