The Q-priming part is really interesting. Making models 5x more likely to ask for clarification instead of guessing wrong is exactly what we need for agentic workflows. Right now most agents just confidently charge ahead with bad assumptions.

→ View original post on X — @whats_ai