“Paul believes that building workarounds for the limitations of current AI models is a flawed strategy, as these limitations are likely to be resolved in subsequent model iterations.”