AI judgment may follow a scaling law | Scaling AI’s judgment of usability heuristics | Generating consistent visual design assets with AI | New AI music model: Mureka O2
The scaling law framing is sharp, but I think the most useful read of your data is as a sorting function, not just a capability curve.
I run creative organizations. For decades, the B+ creative director survived on process fluency and proximity to good work. Right references, confident delivery, strategies reverse-engineered from other people's case studies. They performed depth. AI just gave everyone that same performance for free.
What's interesting is that the A-level creatives didn't lose ground. They gained it. Because what scales with compute is pattern-matching and synthesis.
What doesn't scale is having been wrong in rooms where it cost you something, or having sat with a brief for three days and come back with something that made the strategy team uncomfortable.
Your Baymard example is perfect for this. AI went from 39 guidelines to 154 in eight months. That's the performed-depth layer getting automated. The judgment layer underneath, the one that knows which heuristic matters most for this user in this context, that's the part where your hoped-for scaling law will face the real test.
AI isn't creating two tiers of professionals. It's revealing two tiers that were always there.
The scaling law framing is sharp, but I think the most useful read of your data is as a sorting function, not just a capability curve.
I run creative organizations. For decades, the B+ creative director survived on process fluency and proximity to good work. Right references, confident delivery, strategies reverse-engineered from other people's case studies. They performed depth. AI just gave everyone that same performance for free.
What's interesting is that the A-level creatives didn't lose ground. They gained it. Because what scales with compute is pattern-matching and synthesis.
What doesn't scale is having been wrong in rooms where it cost you something, or having sat with a brief for three days and come back with something that made the strategy team uncomfortable.
Your Baymard example is perfect for this. AI went from 39 guidelines to 154 in eight months. That's the performed-depth layer getting automated. The judgment layer underneath, the one that knows which heuristic matters most for this user in this context, that's the part where your hoped-for scaling law will face the real test.
AI isn't creating two tiers of professionals. It's revealing two tiers that were always there.
Hey, great read as always. 'Meatware supremacy' – ha! So accurate.