TL;DR
The short version
The "AI kills PMs" vibe flipped. Anthropic's Ami Vora and Mike Krieger say the job is back — and harder: bridge humans to fast-changing models, convene work Claude can't, and ship agent-native software without losing coherence.
From the Lenny & Friends Summit panel: glue intensified, pixel-perfect prediction depreciated, bet leads matter, and the mission is closing the gap between model capability and what most people use.
Ami Vora (CPO) and Mike Krieger (Head of Labs), moderated by Dan Shipper at Lenny & Friends Summit, 10 Sep 2026; published on Lenny's Podcast 29 Sep 2026.
Humans stay slow; models do not
Vora's frame is durable: product bridges human problems to technology. Cadence broke — tech can feel like it shifts every two months inside a lab while human needs do not. Krieger's story makes the ops half concrete: Ami pulled him from pure building into glue work that was about to drop. Moving faster raised the bar for operational excellence.
I don't think Claude is a convenor yet. Somebody still needs to bring the Claudes and the people together.
Mike Krieger
Old craft expires; adaptation does not
Pixel-level product/design prediction depreciates when you can ship three variants. The appreciating asset is constant relearning — and framing chaos so the team is not taxed into paralysis. Lab practice: explore widely, name bet leads / DRIs who decide whether a frontier thread continues.
Agent-native, then fold back
Krieger's thesis: anything a human can do, an agent should be able to do. Internally, much Anthropic software is already Claude-built and iterated. For scaled SaaS: evolve surfaces on shared primitives, run overlapping experiments until something works, then fold back — invest in foundations so the third experiment is not born without memory.
The closer is the durable mission: close the gap between what models can do and what most people use. That is still product work.
