Models are still getting better fast, and an architecture that makes sense today may not in a few months. More worth remembering than any particular structure is the principle: measure first, change one thing at a time, keep what survives testing.

"Models matter a lot, but by themselves, they don't solve a problem. It's the entire end-to-end solution." Marc Boroditsky, CRO of @nebiusai , joined Mattie Toia, VP of Infrastructure at @Uber, and Dion Harris, Sr. Director of HPC & AI Hyperscale Infrastructure Solutions at @nvidia, to discuss why improving compute economics takes the whole AI ecosystem, not any single company. More from our AI Agenda Live event:

Models keep getting smarter. But the harness doesn’t disappear. Instead, its job shifts from making decisions for the model to giving it options. In Replit Agent, the main agent decides when to think harder, when to hand work to a subagent, and how much effort that subagent should spend.