33:35

Luca Mezzalira and Teena Idnani open with a clear framing: AI is accelerating certain tasks architects do, but the fundamentals — understanding context, connecting technical decisions to business capability, designing for evolutionary systems — are unchanged and arguably more important than ever. Luca identifies what he calls the shift from a T-shaped to an "M-shaped" architect: broader and deeper simultaneously, using AI as a research accelerator that collapses days of trade-off analysis into hours. Both agree that the biggest current risk isn't that AI will replace architects, but that AI-generated code can look convincingly correct and pass initial testing while concealing edge-case failures that only surface under load, regulatory audit, or upstream change — particularly dangerous in regulated industries like finance. The conversation sharpens around what architects must double down on to stay relevant. Luca advocates for merging deterministic and probabilistic systems through harness engineering — combining linters, static analysis, and deterministic guardrails with AI code assistants to produce more predictable outcomes. More broadly, both conclude that empathy is now a core technical skill: the architect's real job is riding the "elevator" between the engine room and the C-suite, translating freely in both directions. Luca's rule of thumb has shifted from 70% people and context, 30% technical — to nearly 90/10 in the AI era. The competitive edge for architects, they agree, is not in generating code, but in knowing which problems are worth solving and why.