An early set from the desk

My point exactly, cleanly put: the edge in enterprise AI doesn’t come from the best foundation model but from the “operating layer” — AI embedded in operations, where proprietary data, domain experts and feedback loops become the real moat. Pilot-to-P&L, rendered as architecture.

Ulrich Homann (Microsoft) on the shift from coder to “Product Engineer”: as AI cheapens implementation, value moves to architecture and problem understanding. My addition from building things myself — the tool lowers the barrier to building, not the barrier to building well.

Anthropic’s own data: engineers ship 8× more code per quarter, over 80 % authored by Claude. The view on the trajectory behind “vibe coding” — and why Industrial AI readiness is not a someday-luxury but a today-necessity.

AI readiness isn’t only a question for schools. The WEF framework reads as a blueprint for organisations too: taking AI seriously means designing for people’s ability to learn and adapt — not just the models.

Wermke, Bomke and Kerkmann deliver the update to last year’s “decisive year” framing: twelve months on, Physical AI is no longer a slogan but a production reality. The Pilot-to-P&L question stays open.