What's this about?
Mollick explains his method: he posts screenshots and artifacts instead of links because nobody clicks links — visual output is the only thing that communicates model progress at feed speed. The rigorous research lives at his Penn lab.
What exactly is the trade?
Attention economics: a benchmark table gets scrolled past; a rendered simulation stops the scroll. The demo *is* the argument — with the obvious risk that impressive-looking beats actually-good.
Why is this worth noting?
It's the candid mechanics behind how AI capability perception forms: the most-shared evidence is optimized for scroll-stopping, by explicit strategy — worth keeping in mind about every viral demo, including the ones in today's digest.